Regression test for issue #2175.

Ref: issue #2175

.gitignore

@@ -19,6 +19,7 @@ src/transfer.sh
 src/configs
 redis.ds
 src/redis.conf
+src/nodes.conf
 deps/lua/src/lua
 deps/lua/src/luac
 deps/lua/src/liblua.a

00-RELEASENOTES

@@ -5,7 +5,7 @@ There is no release notes for this branch, it gets forked into another branch
 every time there is a partial feature freeze in order to eventually create
 a new stable release.
 
-Usually "unstable" is stable enough for you to use it in development enviromnets
+Usually "unstable" is stable enough for you to use it in development environments
 however you should never use it in production environments. It is possible
 to download the latest stable release here:
 

BUGS

@@ -1 +1 @@
-Plese check https://github.com/antirez/redis/issues
+Please check https://github.com/antirez/redis/issues

deps/Makefile

@@ -58,12 +58,17 @@ ifeq ($(uname_S),SunOS)
 	LUA_CFLAGS= -D__C99FEATURES__=1
 endif
 
-LUA_CFLAGS+= -O2 -Wall -DLUA_ANSI $(CFLAGS)
+LUA_CFLAGS+= -O2 -Wall -DLUA_ANSI -DENABLE_CJSON_GLOBAL $(CFLAGS)
 LUA_LDFLAGS+= $(LDFLAGS)
+# lua's Makefile defines AR="ar rcu", which is unusual, and makes it more
+# challenging to cross-compile lua (and redis).  These defines make it easier
+# to fit redis into cross-compilation environments, which typically set AR.
+AR=ar
+ARFLAGS=rcu
 
 lua: .make-prerequisites
 	@printf '%b %b\n' $(MAKECOLOR)MAKE$(ENDCOLOR) $(BINCOLOR)$@$(ENDCOLOR)
-	cd lua/src && $(MAKE) all CFLAGS="$(LUA_CFLAGS)" MYLDFLAGS="$(LUA_LDFLAGS)"
+	cd lua/src && $(MAKE) all CFLAGS="$(LUA_CFLAGS)" MYLDFLAGS="$(LUA_LDFLAGS)" AR="$(AR) $(ARFLAGS)"
 
 .PHONY: lua
 

deps/hiredis/fmacros.h

@@ -5,6 +5,10 @@
 #define _BSD_SOURCE
 #endif
 
+#if defined(_AIX)
+#define _ALL_SOURCE
+#endif
+
 #if defined(__sun__)
 #define _POSIX_C_SOURCE 200112L
 #elif defined(__linux__) || defined(__OpenBSD__) || defined(__NetBSD__)

deps/hiredis/net.c

@@ -300,6 +300,7 @@ static int _redisContextConnectTcp(redisContext *c, const char *addr, int port,
                     break;
                 }
             }
+            freeaddrinfo(bservinfo);
             if (!bound) {
                 char buf[128];
                 snprintf(buf,sizeof(buf),"Can't bind socket: %s",strerror(errno));

deps/hiredis/net.h

@@ -35,7 +35,7 @@
 
 #include "hiredis.h"
 
-#if defined(__sun)
+#if defined(__sun) || defined(_AIX)
 #define AF_LOCAL AF_UNIX
 #endif
 

deps/hiredis/sds.c

@@ -123,7 +123,7 @@ void sdsclear(sds s) {
 /* Enlarge the free space at the end of the sds string so that the caller
  * is sure that after calling this function can overwrite up to addlen
  * bytes after the end of the string, plus one more byte for nul term.
- * 
+ *
  * Note: this does not change the *length* of the sds string as returned
  * by sdslen(), but only the free buffer space we have. */
 sds sdsMakeRoomFor(sds s, size_t addlen) {
@@ -200,10 +200,12 @@ size_t sdsAllocSize(sds s) {
 void sdsIncrLen(sds s, int incr) {
     struct sdshdr *sh = (void*) (s-(sizeof(struct sdshdr)));
 
-    assert(sh->free >= incr);
+    if (incr >= 0)
+        assert(sh->free >= (unsigned int)incr);
+    else
+        assert(sh->len >= (unsigned int)(-incr));
     sh->len += incr;
     sh->free -= incr;
-    assert(sh->free >= 0);
     s[sh->len] = '\0';
 }
 
@@ -388,6 +390,7 @@ sds sdscatvprintf(sds s, const char *fmt, va_list ap) {
         buf[buflen-2] = '\0';
         va_copy(cpy,ap);
         vsnprintf(buf, buflen, fmt, cpy);
+        va_end(ap);
         if (buf[buflen-2] != '\0') {
             if (buf != staticbuf) zfree(buf);
             buflen *= 2;
@@ -457,7 +460,7 @@ sds sdscatfmt(sds s, char const *fmt, ...) {
     i = initlen; /* Position of the next byte to write to dest str. */
     while(*f) {
         char next, *str;
-        int l;
+        unsigned int l;
         long long num;
         unsigned long long unum;
 

deps/hiredis/sds.h

@@ -39,8 +39,8 @@
 typedef char *sds;
 
 struct sdshdr {
-    int len;
-    int free;
+    unsigned int len;
+    unsigned int free;
     char buf[];
 };
 

deps/hiredis/test.c

@@ -51,7 +51,7 @@ static redisContext *select_database(redisContext *c) {
     assert(reply != NULL);
     freeReplyObject(reply);
 
-    /* Make sure the DB is emtpy */
+    /* Make sure the DB is empty */
     reply = redisCommand(c,"DBSIZE");
     assert(reply != NULL);
     if (reply->type == REDIS_REPLY_INTEGER && reply->integer == 0) {

deps/linenoise/.gitignore

@@ -1 +1,3 @@
-linenoise_example*
+linenoise_example
+*.dSYM
+history.txt

deps/linenoise/README.markdown

@@ -1,8 +1,13 @@
 # Linenoise
 
-A minimal, zero-config, BSD licensed, readline replacement.
+A minimal, zero-config, BSD licensed, readline replacement used in Redis,
+MongoDB, and Android.
 
-News: linenoise is now part of [Android](http://android.git.kernel.org/?p=platform/system/core.git;a=tree;f=liblinenoise;h=56450eaed7f783760e5e6a5993ef75cde2e29dea;hb=HEAD Android)!
+* Single and multi line editing mode with the usual key bindings implemented.
+* History handling.
+* Completion.
+* About 1,100 lines of BSD license source code.
+* Only uses a subset of VT100 escapes (ANSI.SYS compatible).
 
 ## Can a line editing library be 20k lines of code?
 
@@ -10,7 +15,7 @@ Line editing with some support for history is a really important feature for com
 
 So what usually happens is either:
 
- * Large programs with configure scripts disabling line editing if readline is not present in the system, or not supporting it at all since readline is GPL licensed and libedit (the BSD clone) is not as known and available as readline is (Readl world example of this problem: Tclsh).
+ * Large programs with configure scripts disabling line editing if readline is not present in the system, or not supporting it at all since readline is GPL licensed and libedit (the BSD clone) is not as known and available as readline is (Real world example of this problem: Tclsh).
  * Smaller programs not using a configure script not supporting line editing at all (A problem we had with Redis-cli for instance).
  
 The result is a pollution of binaries without line editing support.
@@ -19,27 +24,29 @@ So I spent more or less two hours doing a reality check resulting in this little
 
 ## Terminals, in 2010.
 
-Apparently almost every terminal you can happen to use today has some kind of support for VT100 alike escape sequences. So I tried to write a lib using just very basic VT100 features. The resulting library appears to work everywhere I tried to use it.
+Apparently almost every terminal you can happen to use today has some kind of support for basic VT100 escape sequences. So I tried to write a lib using just very basic VT100 features. The resulting library appears to work everywhere I tried to use it, and now can work even on ANSI.SYS compatible terminals, since no
+VT220 specific sequences are used anymore.
 
-Since it's so young I guess there are a few bugs, or the lib may not compile or work with some operating system, but it's a matter of a few weeks and eventually we'll get it right, and there will be no excuses for not shipping command line tools without built-in line editing support.
-
-The library is currently less than 400 lines of code. In order to use it in your project just look at the *example.c* file in the source distribution, it is trivial. Linenoise is BSD code, so you can use both in free software and commercial software.
+The library is currently about 1100 lines of code. In order to use it in your project just look at the *example.c* file in the source distribution, it is trivial. Linenoise is BSD code, so you can use both in free software and commercial software.
 
 ## Tested with...
 
  * Linux text only console ($TERM = linux)
  * Linux KDE terminal application ($TERM = xterm)
  * Linux xterm ($TERM = xterm)
+ * Linux Buildroot ($TERM = vt100)
  * Mac OS X iTerm ($TERM = xterm)
  * Mac OS X default Terminal.app ($TERM = xterm)
  * OpenBSD 4.5 through an OSX Terminal.app ($TERM = screen)
  * IBM AIX 6.1
  * FreeBSD xterm ($TERM = xterm)
+ * ANSI.SYS
 
 Please test it everywhere you can and report back!
 
 ## Let's push this forward!
 
-Please fork it and add something interesting and send me a pull request. What's especially interesting are fixes, new key bindings, completion.
+Patches should be provided in the respect of linenoise sensibility for small
+easy to understand code.
 
 Send feedbacks to antirez at gmail

deps/linenoise/example.c

@@ -1,5 +1,6 @@
 #include <stdio.h>
 #include <stdlib.h>
+#include <string.h>
 #include "linenoise.h"
 
 
@@ -10,16 +11,52 @@ void completion(const char *buf, linenoiseCompletions *lc) {
     }
 }
 
-int main(void) {
+int main(int argc, char **argv) {
     char *line;
+    char *prgname = argv[0];
 
+    /* Parse options, with --multiline we enable multi line editing. */
+    while(argc > 1) {
+        argc--;
+        argv++;
+        if (!strcmp(*argv,"--multiline")) {
+            linenoiseSetMultiLine(1);
+            printf("Multi-line mode enabled.\n");
+        } else if (!strcmp(*argv,"--keycodes")) {
+            linenoisePrintKeyCodes();
+            exit(0);
+        } else {
+            fprintf(stderr, "Usage: %s [--multiline] [--keycodes]\n", prgname);
+            exit(1);
+        }
+    }
+
+    /* Set the completion callback. This will be called every time the
+     * user uses the <tab> key. */
     linenoiseSetCompletionCallback(completion);
+
+    /* Load history from file. The history file is just a plain text file
+     * where entries are separated by newlines. */
     linenoiseHistoryLoad("history.txt"); /* Load the history at startup */
+
+    /* Now this is the main loop of the typical linenoise-based application.
+     * The call to linenoise() will block as long as the user types something
+     * and presses enter.
+     *
+     * The typed string is returned as a malloc() allocated string by
+     * linenoise, so the user needs to free() it. */
     while((line = linenoise("hello> ")) != NULL) {
-        if (line[0] != '\0') {
+        /* Do something with the string. */
+        if (line[0] != '\0' && line[0] != '/') {
             printf("echo: '%s'\n", line);
-            linenoiseHistoryAdd(line);
-            linenoiseHistorySave("history.txt"); /* Save every new entry */
+            linenoiseHistoryAdd(line); /* Add to the history. */
+            linenoiseHistorySave("history.txt"); /* Save the history on disk. */
+        } else if (!strncmp(line,"/historylen",11)) {
+            /* The "/historylen" command will change the history len. */
+            int len = atoi(line+11);
+            linenoiseHistorySetMaxLen(len);
+        } else if (line[0] == '/') {
+            printf("Unreconized command: %s\n", line);
         }
         free(line);
     }

deps/linenoise/linenoise.c

@@ -2,7 +2,7 @@
  * line editing lib needs to be 20,000 lines of C code.
  *
  * You can find the latest source code at:
- * 
+ *
  *   http://github.com/antirez/linenoise
  *
  * Does a number of crazy assumptions that happen to be true in 99.9999% of
@@ -14,18 +14,18 @@
  * Copyright (c) 2010-2013, Pieter Noordhuis <pcnoordhuis at gmail dot com>
  *
  * All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are
  * met:
- * 
+ *
  *  *  Redistributions of source code must retain the above copyright
  *     notice, this list of conditions and the following disclaimer.
  *
  *  *  Redistributions in binary form must reproduce the above copyright
  *     notice, this list of conditions and the following disclaimer in the
  *     documentation and/or other materials provided with the distribution.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
@@ -37,7 +37,7 @@
  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- * 
+ *
  * ------------------------------------------------------------------------
  *
  * References:
@@ -56,10 +56,6 @@
  * flickering effect with some slow terminal, but the lesser sequences
  * the more compatible.
  *
- * CHA (Cursor Horizontal Absolute)
- *    Sequence: ESC [ n G
- *    Effect: moves cursor to column n
- *
  * EL (Erase Line)
  *    Sequence: ESC [ n K
  *    Effect: if n is 0 or missing, clear from cursor to end of line
@@ -68,7 +64,19 @@
  *
  * CUF (CUrsor Forward)
  *    Sequence: ESC [ n C
- *    Effect: moves cursor forward of n chars
+ *    Effect: moves cursor forward n chars
+ *
+ * CUB (CUrsor Backward)
+ *    Sequence: ESC [ n D
+ *    Effect: moves cursor backward n chars
+ *
+ * The following is used to get the terminal width if getting
+ * the width with the TIOCGWINSZ ioctl fails
+ *
+ * DSR (Device Status Report)
+ *    Sequence: ESC [ 6 n
+ *    Effect: reports the current cusor position as ESC [ n ; m R
+ *            where n is the row and m is the column
  *
  * When multi line mode is enabled, we also use an additional escape
  * sequence. However multi line editing is disabled by default.
@@ -81,17 +89,18 @@
  *    Sequence: ESC [ n B
  *    Effect: moves cursor down of n chars.
  *
- * The following are used to clear the screen: ESC [ H ESC [ 2 J
- * This is actually composed of two sequences:
+ * When linenoiseClearScreen() is called, two additional escape sequences
+ * are used in order to clear the screen and position the cursor at home
+ * position.
  *
- * cursorhome
+ * CUP (Cursor position)
  *    Sequence: ESC [ H
  *    Effect: moves the cursor to upper left corner
  *
- * ED2 (Clear entire screen)
+ * ED (Erase display)
  *    Sequence: ESC [ 2 J
  *    Effect: clear the whole screen
- * 
+ *
  */
 
 #include <termios.h>
@@ -332,7 +341,7 @@ static void freeCompletions(linenoiseCompletions *lc) {
 /* This is an helper function for linenoiseEdit() and is called when the
  * user types the <tab> key in order to complete the string currently in the
  * input.
- * 
+ *
  * The state of the editing is encapsulated into the pointed linenoiseState
  * structure as described in the structure definition. */
 static int completeLine(struct linenoiseState *ls) {
@@ -459,7 +468,7 @@ static void refreshSingleLine(struct linenoiseState *l) {
     size_t len = l->len;
     size_t pos = l->pos;
     struct abuf ab;
-    
+
     while((plen+pos) >= l->cols) {
         buf++;
         len--;
@@ -471,7 +480,7 @@ static void refreshSingleLine(struct linenoiseState *l) {
 
     abInit(&ab);
     /* Cursor to left edge */
-    snprintf(seq,64,"\x1b[0G");
+    snprintf(seq,64,"\r");
     abAppend(&ab,seq,strlen(seq));
     /* Write the prompt and the current buffer content */
     abAppend(&ab,l->prompt,strlen(l->prompt));
@@ -480,7 +489,7 @@ static void refreshSingleLine(struct linenoiseState *l) {
     snprintf(seq,64,"\x1b[0K");
     abAppend(&ab,seq,strlen(seq));
     /* Move cursor to original position. */
-    snprintf(seq,64,"\x1b[0G\x1b[%dC", (int)(pos+plen));
+    snprintf(seq,64,"\r\x1b[%dC", (int)(pos+plen));
     abAppend(&ab,seq,strlen(seq));
     if (write(fd,ab.b,ab.len) == -1) {} /* Can't recover from write error. */
     abFree(&ab);
@@ -496,6 +505,7 @@ static void refreshMultiLine(struct linenoiseState *l) {
     int rows = (plen+l->len+l->cols-1)/l->cols; /* rows used by current buf. */
     int rpos = (plen+l->oldpos+l->cols)/l->cols; /* cursor relative row. */
     int rpos2; /* rpos after refresh. */
+    int col; /* colum position, zero-based. */
     int old_rows = l->maxrows;
     int fd = l->ofd, j;
     struct abuf ab;
@@ -515,15 +525,15 @@ static void refreshMultiLine(struct linenoiseState *l) {
     /* Now for every row clear it, go up. */
     for (j = 0; j < old_rows-1; j++) {
         lndebug("clear+up");
-        snprintf(seq,64,"\x1b[0G\x1b[0K\x1b[1A");
+        snprintf(seq,64,"\r\x1b[0K\x1b[1A");
         abAppend(&ab,seq,strlen(seq));
     }
 
     /* Clean the top line. */
     lndebug("clear");
-    snprintf(seq,64,"\x1b[0G\x1b[0K");
+    snprintf(seq,64,"\r\x1b[0K");
     abAppend(&ab,seq,strlen(seq));
-    
+
     /* Write the prompt and the current buffer content */
     abAppend(&ab,l->prompt,strlen(l->prompt));
     abAppend(&ab,l->buf,l->len);
@@ -536,7 +546,7 @@ static void refreshMultiLine(struct linenoiseState *l) {
     {
         lndebug("<newline>");
         abAppend(&ab,"\n",1);
-        snprintf(seq,64,"\x1b[0G");
+        snprintf(seq,64,"\r");
         abAppend(&ab,seq,strlen(seq));
         rows++;
         if (rows > (int)l->maxrows) l->maxrows = rows;
@@ -554,8 +564,12 @@ static void refreshMultiLine(struct linenoiseState *l) {
     }
 
     /* Set column. */
-    lndebug("set col %d", 1+((plen+(int)l->pos) % (int)l->cols));
-    snprintf(seq,64,"\x1b[%dG", 1+((plen+(int)l->pos) % (int)l->cols));
+    col = (plen+(int)l->pos) % (int)l->cols;
+    lndebug("set col %d", 1+col);
+    if (col)
+        snprintf(seq,64,"\r\x1b[%dC", col);
+    else
+        snprintf(seq,64,"\r");
     abAppend(&ab,seq,strlen(seq));
 
     lndebug("\n");
@@ -732,7 +746,7 @@ static int linenoiseEdit(int stdin_fd, int stdout_fd, char *buf, size_t buflen,
     /* The latest history entry is always our current buffer, that
      * initially is just an empty string. */
     linenoiseHistoryAdd("");
-    
+
     if (write(l.ofd,prompt,l.plen) == -1) return -1;
     while(1) {
         char c;
@@ -757,6 +771,7 @@ static int linenoiseEdit(int stdin_fd, int stdout_fd, char *buf, size_t buflen,
         case ENTER:    /* enter */
             history_len--;
             free(history[history_len]);
+            if (mlmode) linenoiseEditMoveEnd(&l);
             return (int)l.len;
         case CTRL_C:     /* ctrl-c */
             errno = EAGAIN;
@@ -765,8 +780,8 @@ static int linenoiseEdit(int stdin_fd, int stdout_fd, char *buf, size_t buflen,
         case 8:     /* ctrl-h */
             linenoiseEditBackspace(&l);
             break;
-        case CTRL_D:     /* ctrl-d, remove char at right of cursor, or of the
-                       line is empty, act as end-of-file. */
+        case CTRL_D:     /* ctrl-d, remove char at right of cursor, or if the
+                            line is empty, act as end-of-file. */
             if (l.len > 0) {
                 linenoiseEditDelete(&l);
             } else {
@@ -904,7 +919,7 @@ void linenoisePrintKeyCodes(void) {
 
         printf("'%c' %02x (%d) (type quit to exit)\n",
             isprint(c) ? c : '?', (int)c, (int)c);
-        printf("\x1b[0G"); /* Go left edge manually, we are in raw mode. */
+        printf("\r"); /* Go left edge manually, we are in raw mode. */
         fflush(stdout);
     }
     disableRawMode(STDIN_FILENO);
@@ -1058,7 +1073,7 @@ int linenoiseHistorySetMaxLen(int len) {
 int linenoiseHistorySave(const char *filename) {
     FILE *fp = fopen(filename,"w");
     int j;
-    
+
     if (fp == NULL) return -1;
     for (j = 0; j < history_len; j++)
         fprintf(fp,"%s\n",history[j]);
@@ -1074,12 +1089,12 @@ int linenoiseHistorySave(const char *filename) {
 int linenoiseHistoryLoad(const char *filename) {
     FILE *fp = fopen(filename,"r");
     char buf[LINENOISE_MAX_LINE];
-    
+
     if (fp == NULL) return -1;
 
     while (fgets(buf,LINENOISE_MAX_LINE,fp) != NULL) {
         char *p;
-        
+
         p = strchr(buf,'\r');
         if (!p) p = strchr(buf,'\n');
         if (p) *p = '\0';

deps/linenoise/linenoise.h

@@ -9,18 +9,18 @@
  * Copyright (c) 2010, Pieter Noordhuis <pcnoordhuis at gmail dot com>
  *
  * All rights reserved.
- * 
+ *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are
  * met:
- * 
+ *
  *  *  Redistributions of source code must retain the above copyright
  *     notice, this list of conditions and the following disclaimer.
  *
  *  *  Redistributions in binary form must reproduce the above copyright
  *     notice, this list of conditions and the following disclaimer in the
  *     documentation and/or other materials provided with the distribution.
- * 
+ *
  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR

deps/lua/src/Makefile

@@ -25,9 +25,10 @@ PLATS= aix ansi bsd freebsd generic linux macosx mingw posix solaris
 LUA_A=	liblua.a
 CORE_O=	lapi.o lcode.o ldebug.o ldo.o ldump.o lfunc.o lgc.o llex.o lmem.o \
 	lobject.o lopcodes.o lparser.o lstate.o lstring.o ltable.o ltm.o  \
-	lundump.o lvm.o lzio.o strbuf.o
+	lundump.o lvm.o lzio.o strbuf.o fpconv.o
 LIB_O=	lauxlib.o lbaselib.o ldblib.o liolib.o lmathlib.o loslib.o ltablib.o \
-	lstrlib.o loadlib.o linit.o lua_cjson.o lua_struct.o lua_cmsgpack.o
+	lstrlib.o loadlib.o linit.o lua_cjson.o lua_struct.o lua_cmsgpack.o \
+	lua_bit.o
 
 LUA_T=	lua
 LUA_O=	lua.o

deps/lua/src/fpconv.c

@@ -0,0 +1,205 @@
+/* fpconv - Floating point conversion routines
+ *
+ * Copyright (c) 2011-2012  Mark Pulford <mark@kyne.com.au>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
+ * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/* JSON uses a '.' decimal separator. strtod() / sprintf() under C libraries
+ * with locale support will break when the decimal separator is a comma.
+ *
+ * fpconv_* will around these issues with a translation buffer if required.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <assert.h>
+#include <string.h>
+
+#include "fpconv.h"
+
+/* Lua CJSON assumes the locale is the same for all threads within a
+ * process and doesn't change after initialisation.
+ *
+ * This avoids the need for per thread storage or expensive checks
+ * for call. */
+static char locale_decimal_point = '.';
+
+/* In theory multibyte decimal_points are possible, but
+ * Lua CJSON only supports UTF-8 and known locales only have
+ * single byte decimal points ([.,]).
+ *
+ * localconv() may not be thread safe (=>crash), and nl_langinfo() is
+ * not supported on some platforms. Use sprintf() instead - if the
+ * locale does change, at least Lua CJSON won't crash. */
+static void fpconv_update_locale()
+{
+    char buf[8];
+
+    snprintf(buf, sizeof(buf), "%g", 0.5);
+
+    /* Failing this test might imply the platform has a buggy dtoa
+     * implementation or wide characters */
+    if (buf[0] != '0' || buf[2] != '5' || buf[3] != 0) {
+        fprintf(stderr, "Error: wide characters found or printf() bug.");
+        abort();
+    }
+
+    locale_decimal_point = buf[1];
+}
+
+/* Check for a valid number character: [-+0-9a-yA-Y.]
+ * Eg: -0.6e+5, infinity, 0xF0.F0pF0
+ *
+ * Used to find the probable end of a number. It doesn't matter if
+ * invalid characters are counted - strtod() will find the valid
+ * number if it exists.  The risk is that slightly more memory might
+ * be allocated before a parse error occurs. */
+static inline int valid_number_character(char ch)
+{
+    char lower_ch;
+
+    if ('0' <= ch && ch <= '9')
+        return 1;
+    if (ch == '-' || ch == '+' || ch == '.')
+        return 1;
+
+    /* Hex digits, exponent (e), base (p), "infinity",.. */
+    lower_ch = ch | 0x20;
+    if ('a' <= lower_ch && lower_ch <= 'y')
+        return 1;
+
+    return 0;
+}
+
+/* Calculate the size of the buffer required for a strtod locale
+ * conversion. */
+static int strtod_buffer_size(const char *s)
+{
+    const char *p = s;
+
+    while (valid_number_character(*p))
+        p++;
+
+    return p - s;
+}
+
+/* Similar to strtod(), but must be passed the current locale's decimal point
+ * character. Guaranteed to be called at the start of any valid number in a string */
+double fpconv_strtod(const char *nptr, char **endptr)
+{
+    char localbuf[FPCONV_G_FMT_BUFSIZE];
+    char *buf, *endbuf, *dp;
+    int buflen;
+    double value;
+
+    /* System strtod() is fine when decimal point is '.' */
+    if (locale_decimal_point == '.')
+        return strtod(nptr, endptr);
+
+    buflen = strtod_buffer_size(nptr);
+    if (!buflen) {
+        /* No valid characters found, standard strtod() return */
+        *endptr = (char *)nptr;
+        return 0;
+    }
+
+    /* Duplicate number into buffer */
+    if (buflen >= FPCONV_G_FMT_BUFSIZE) {
+        /* Handle unusually large numbers */
+        buf = malloc(buflen + 1);
+        if (!buf) {
+            fprintf(stderr, "Out of memory");
+            abort();
+        }
+    } else {
+        /* This is the common case.. */
+        buf = localbuf;
+    }
+    memcpy(buf, nptr, buflen);
+    buf[buflen] = 0;
+
+    /* Update decimal point character if found */
+    dp = strchr(buf, '.');
+    if (dp)
+        *dp = locale_decimal_point;
+
+    value = strtod(buf, &endbuf);
+    *endptr = (char *)&nptr[endbuf - buf];
+    if (buflen >= FPCONV_G_FMT_BUFSIZE)
+        free(buf);
+
+    return value;
+}
+
+/* "fmt" must point to a buffer of at least 6 characters */
+static void set_number_format(char *fmt, int precision)
+{
+    int d1, d2, i;
+
+    assert(1 <= precision && precision <= 14);
+
+    /* Create printf format (%.14g) from precision */
+    d1 = precision / 10;
+    d2 = precision % 10;
+    fmt[0] = '%';
+    fmt[1] = '.';
+    i = 2;
+    if (d1) {
+        fmt[i++] = '0' + d1;
+    }
+    fmt[i++] = '0' + d2;
+    fmt[i++] = 'g';
+    fmt[i] = 0;
+}
+
+/* Assumes there is always at least 32 characters available in the target buffer */
+int fpconv_g_fmt(char *str, double num, int precision)
+{
+    char buf[FPCONV_G_FMT_BUFSIZE];
+    char fmt[6];
+    int len;
+    char *b;
+
+    set_number_format(fmt, precision);
+
+    /* Pass through when decimal point character is dot. */
+    if (locale_decimal_point == '.')
+        return snprintf(str, FPCONV_G_FMT_BUFSIZE, fmt, num);
+
+    /* snprintf() to a buffer then translate for other decimal point characters */
+    len = snprintf(buf, FPCONV_G_FMT_BUFSIZE, fmt, num);
+
+    /* Copy into target location. Translate decimal point if required */
+    b = buf;
+    do {
+        *str++ = (*b == locale_decimal_point ? '.' : *b);
+    } while(*b++);
+
+    return len;
+}
+
+void fpconv_init()
+{
+    fpconv_update_locale();
+}
+
+/* vi:ai et sw=4 ts=4:
+ */

deps/lua/src/fpconv.h

@@ -0,0 +1,22 @@
+/* Lua CJSON floating point conversion routines */
+
+/* Buffer required to store the largest string representation of a double.
+ *
+ * Longest double printed with %.14g is 21 characters long:
+ * -1.7976931348623e+308 */
+# define FPCONV_G_FMT_BUFSIZE   32
+
+#ifdef USE_INTERNAL_FPCONV
+static inline void fpconv_init()
+{
+    /* Do nothing - not required */
+}
+#else
+extern void fpconv_init();
+#endif
+
+extern int fpconv_g_fmt(char*, double, int);
+extern double fpconv_strtod(const char*, char**);
+
+/* vi:ai et sw=4 ts=4:
+ */

deps/lua/src/lua_bit.c

@@ -0,0 +1,189 @@
+/*
+** Lua BitOp -- a bit operations library for Lua 5.1/5.2.
+** http://bitop.luajit.org/
+**
+** Copyright (C) 2008-2012 Mike Pall. All rights reserved.
+**
+** Permission is hereby granted, free of charge, to any person obtaining
+** a copy of this software and associated documentation files (the
+** "Software"), to deal in the Software without restriction, including
+** without limitation the rights to use, copy, modify, merge, publish,
+** distribute, sublicense, and/or sell copies of the Software, and to
+** permit persons to whom the Software is furnished to do so, subject to
+** the following conditions:
+**
+** The above copyright notice and this permission notice shall be
+** included in all copies or substantial portions of the Software.
+**
+** THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+** EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+** MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+** IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
+** CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+** TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+** SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+**
+** [ MIT license: http://www.opensource.org/licenses/mit-license.php ]
+*/
+
+#define LUA_BITOP_VERSION	"1.0.2"
+
+#define LUA_LIB
+#include "lua.h"
+#include "lauxlib.h"
+
+#ifdef _MSC_VER
+/* MSVC is stuck in the last century and doesn't have C99's stdint.h. */
+typedef __int32 int32_t;
+typedef unsigned __int32 uint32_t;
+typedef unsigned __int64 uint64_t;
+#else
+#include <stdint.h>
+#endif
+
+typedef int32_t SBits;
+typedef uint32_t UBits;
+
+typedef union {
+  lua_Number n;
+#ifdef LUA_NUMBER_DOUBLE
+  uint64_t b;
+#else
+  UBits b;
+#endif
+} BitNum;
+
+/* Convert argument to bit type. */
+static UBits barg(lua_State *L, int idx)
+{
+  BitNum bn;
+  UBits b;
+#if LUA_VERSION_NUM < 502
+  bn.n = lua_tonumber(L, idx);
+#else
+  bn.n = luaL_checknumber(L, idx);
+#endif
+#if defined(LUA_NUMBER_DOUBLE)
+  bn.n += 6755399441055744.0;  /* 2^52+2^51 */
+#ifdef SWAPPED_DOUBLE
+  b = (UBits)(bn.b >> 32);
+#else
+  b = (UBits)bn.b;
+#endif
+#elif defined(LUA_NUMBER_INT) || defined(LUA_NUMBER_LONG) || \
+      defined(LUA_NUMBER_LONGLONG) || defined(LUA_NUMBER_LONG_LONG) || \
+      defined(LUA_NUMBER_LLONG)
+  if (sizeof(UBits) == sizeof(lua_Number))
+    b = bn.b;
+  else
+    b = (UBits)(SBits)bn.n;
+#elif defined(LUA_NUMBER_FLOAT)
+#error "A 'float' lua_Number type is incompatible with this library"
+#else
+#error "Unknown number type, check LUA_NUMBER_* in luaconf.h"
+#endif
+#if LUA_VERSION_NUM < 502
+  if (b == 0 && !lua_isnumber(L, idx)) {
+    luaL_typerror(L, idx, "number");
+  }
+#endif
+  return b;
+}
+
+/* Return bit type. */
+#define BRET(b)  lua_pushnumber(L, (lua_Number)(SBits)(b)); return 1;
+
+static int bit_tobit(lua_State *L) { BRET(barg(L, 1)) }
+static int bit_bnot(lua_State *L) { BRET(~barg(L, 1)) }
+
+#define BIT_OP(func, opr) \
+  static int func(lua_State *L) { int i; UBits b = barg(L, 1); \
+    for (i = lua_gettop(L); i > 1; i--) b opr barg(L, i); BRET(b) }
+BIT_OP(bit_band, &=)
+BIT_OP(bit_bor, |=)
+BIT_OP(bit_bxor, ^=)
+
+#define bshl(b, n)  (b << n)
+#define bshr(b, n)  (b >> n)
+#define bsar(b, n)  ((SBits)b >> n)
+#define brol(b, n)  ((b << n) | (b >> (32-n)))
+#define bror(b, n)  ((b << (32-n)) | (b >> n))
+#define BIT_SH(func, fn) \
+  static int func(lua_State *L) { \
+    UBits b = barg(L, 1); UBits n = barg(L, 2) & 31; BRET(fn(b, n)) }
+BIT_SH(bit_lshift, bshl)
+BIT_SH(bit_rshift, bshr)
+BIT_SH(bit_arshift, bsar)
+BIT_SH(bit_rol, brol)
+BIT_SH(bit_ror, bror)
+
+static int bit_bswap(lua_State *L)
+{
+  UBits b = barg(L, 1);
+  b = (b >> 24) | ((b >> 8) & 0xff00) | ((b & 0xff00) << 8) | (b << 24);
+  BRET(b)
+}
+
+static int bit_tohex(lua_State *L)
+{
+  UBits b = barg(L, 1);
+  SBits n = lua_isnone(L, 2) ? 8 : (SBits)barg(L, 2);
+  const char *hexdigits = "0123456789abcdef";
+  char buf[8];
+  int i;
+  if (n < 0) { n = -n; hexdigits = "0123456789ABCDEF"; }
+  if (n > 8) n = 8;
+  for (i = (int)n; --i >= 0; ) { buf[i] = hexdigits[b & 15]; b >>= 4; }
+  lua_pushlstring(L, buf, (size_t)n);
+  return 1;
+}
+
+static const struct luaL_Reg bit_funcs[] = {
+  { "tobit",	bit_tobit },
+  { "bnot",	bit_bnot },
+  { "band",	bit_band },
+  { "bor",	bit_bor },
+  { "bxor",	bit_bxor },
+  { "lshift",	bit_lshift },
+  { "rshift",	bit_rshift },
+  { "arshift",	bit_arshift },
+  { "rol",	bit_rol },
+  { "ror",	bit_ror },
+  { "bswap",	bit_bswap },
+  { "tohex",	bit_tohex },
+  { NULL, NULL }
+};
+
+/* Signed right-shifts are implementation-defined per C89/C99.
+** But the de facto standard are arithmetic right-shifts on two's
+** complement CPUs. This behaviour is required here, so test for it.
+*/
+#define BAD_SAR		(bsar(-8, 2) != (SBits)-2)
+
+LUALIB_API int luaopen_bit(lua_State *L)
+{
+  UBits b;
+  lua_pushnumber(L, (lua_Number)1437217655L);
+  b = barg(L, -1);
+  if (b != (UBits)1437217655L || BAD_SAR) {  /* Perform a simple self-test. */
+    const char *msg = "compiled with incompatible luaconf.h";
+#ifdef LUA_NUMBER_DOUBLE
+#ifdef _WIN32
+    if (b == (UBits)1610612736L)
+      msg = "use D3DCREATE_FPU_PRESERVE with DirectX";
+#endif
+    if (b == (UBits)1127743488L)
+      msg = "not compiled with SWAPPED_DOUBLE";
+#endif
+    if (BAD_SAR)
+      msg = "arithmetic right-shift broken";
+    luaL_error(L, "bit library self-test failed (%s)", msg);
+  }
+#if LUA_VERSION_NUM < 502
+  luaL_register(L, "bit", bit_funcs);
+#else
+  luaL_newlib(L, bit_funcs);
+#endif
+  return 1;
+}
+

deps/lua/src/lua_cmsgpack.c

@@ -7,14 +7,46 @@
 #include "lua.h"
 #include "lauxlib.h"
 
-#define LUACMSGPACK_VERSION     "lua-cmsgpack 0.3.0"
+#define LUACMSGPACK_NAME        "cmsgpack"
+#define LUACMSGPACK_SAFE_NAME   "cmsgpack_safe"
+#define LUACMSGPACK_VERSION     "lua-cmsgpack 0.4.0"
 #define LUACMSGPACK_COPYRIGHT   "Copyright (C) 2012, Salvatore Sanfilippo"
 #define LUACMSGPACK_DESCRIPTION "MessagePack C implementation for Lua"
 
-#define LUACMSGPACK_MAX_NESTING  16 /* Max tables nesting. */
+/* Allows a preprocessor directive to override MAX_NESTING */
+#ifndef LUACMSGPACK_MAX_NESTING
+    #define LUACMSGPACK_MAX_NESTING  16 /* Max tables nesting. */
+#endif
 
-/* ==============================================================================
- * MessagePack implementation and bindings for Lua 5.1.
+#if (_XOPEN_SOURCE >= 600 || _ISOC99_SOURCE || _POSIX_C_SOURCE >= 200112L)
+    #define IS_FINITE(x) isfinite(x)
+#else
+    #define IS_FINITE(x) ((x) == (x) && (x) + 1 > (x))
+#endif
+
+/* Check if float or double can be an integer without loss of precision */
+#define IS_INT_TYPE_EQUIVALENT(x, T) (IS_FINITE(x) && (T)(x) == (x))
+
+#define IS_INT64_EQUIVALENT(x) IS_INT_TYPE_EQUIVALENT(x, int64_t)
+#define IS_INT_EQUIVALENT(x) IS_INT_TYPE_EQUIVALENT(x, int)
+
+/* If size of pointer is equal to a 4 byte integer, we're on 32 bits. */
+#if UINTPTR_MAX == UINT_MAX
+    #define BITS_32 1
+#else
+    #define BITS_32 0
+#endif
+
+#if LUA_VERSION_NUM < 503
+    #if BITS_32
+        #define lua_pushunsigned(L, n) lua_pushnumber(L, n)
+    #else
+        #define lua_pushunsigned(L, n) lua_pushinteger(L, n)
+    #endif
+#endif
+
+/* =============================================================================
+ * MessagePack implementation and bindings for Lua 5.1/5.2.
  * Copyright(C) 2012 Salvatore Sanfilippo <antirez@gmail.com>
  *
  * http://github.com/antirez/lua-cmsgpack
@@ -29,23 +61,27 @@
  * 20-Feb-2012 (ver 0.2.0): Tables encoding improved.
  * 20-Feb-2012 (ver 0.2.1): Minor bug fixing.
  * 20-Feb-2012 (ver 0.3.0): Module renamed lua-cmsgpack (was lua-msgpack).
- * ============================================================================ */
+ * 04-Apr-2014 (ver 0.3.1): Lua 5.2 support and minor bug fix.
+ * 07-Apr-2014 (ver 0.4.0): Multiple pack/unpack, lua allocator, efficiency.
+ * ========================================================================== */
 
-/* --------------------------- Endian conversion --------------------------------
- * We use it only for floats and doubles, all the other conversions are performed
+/* -------------------------- Endian conversion --------------------------------
+ * We use it only for floats and doubles, all the other conversions performed
  * in an endian independent fashion. So the only thing we need is a function
- * that swaps a binary string if the arch is little endian (and left it untouched
+ * that swaps a binary string if arch is little endian (and left it untouched
  * otherwise). */
 
 /* Reverse memory bytes if arch is little endian. Given the conceptual
- * simplicity of the Lua build system we prefer to check for endianess at runtime.
+ * simplicity of the Lua build system we prefer check for endianess at runtime.
  * The performance difference should be acceptable. */
 static void memrevifle(void *ptr, size_t len) {
-    unsigned char *p = ptr, *e = p+len-1, aux;
+    unsigned char   *p = (unsigned char *)ptr,
+                    *e = (unsigned char *)p+len-1,
+                    aux;
     int test = 1;
     unsigned char *testp = (unsigned char*) &test;
 
-    if (testp[0] == 0) return; /* Big endian, nothign to do. */
+    if (testp[0] == 0) return; /* Big endian, nothing to do. */
     len /= 2;
     while(len--) {
         aux = *p;
@@ -56,30 +92,44 @@ static void memrevifle(void *ptr, size_t len) {
     }
 }
 
-/* ----------------------------- String buffer ----------------------------------
- * This is a simple implementation of string buffers. The only opereation
+/* ---------------------------- String buffer ----------------------------------
+ * This is a simple implementation of string buffers. The only operation
  * supported is creating empty buffers and appending bytes to it.
  * The string buffer uses 2x preallocation on every realloc for O(N) append
  * behavior.  */
 
 typedef struct mp_buf {
+    lua_State *L;
     unsigned char *b;
     size_t len, free;
 } mp_buf;
 
-static mp_buf *mp_buf_new(void) {
-    mp_buf *buf = malloc(sizeof(*buf));
-    
+static void *mp_realloc(lua_State *L, void *target, size_t osize,size_t nsize) {
+    void *(*local_realloc) (void *, void *, size_t osize, size_t nsize) = NULL;
+    void *ud;
+
+    local_realloc = lua_getallocf(L, &ud);
+
+    return local_realloc(ud, target, osize, nsize);
+}
+
+static mp_buf *mp_buf_new(lua_State *L) {
+    mp_buf *buf = NULL;
+
+    /* Old size = 0; new size = sizeof(*buf) */
+    buf = (mp_buf*)mp_realloc(L, NULL, 0, sizeof(*buf));
+
+    buf->L = L;
     buf->b = NULL;
     buf->len = buf->free = 0;
     return buf;
 }
 
-void mp_buf_append(mp_buf *buf, const unsigned char *s, size_t len) {
+static void mp_buf_append(mp_buf *buf, const unsigned char *s, size_t len) {
     if (buf->free < len) {
         size_t newlen = buf->len+len;
 
-        buf->b = realloc(buf->b,newlen*2);
+        buf->b = (unsigned char*)mp_realloc(buf->L, buf->b, buf->len, newlen*2);
         buf->free = newlen;
     }
     memcpy(buf->b+buf->len,s,len);
@@ -88,11 +138,11 @@ void mp_buf_append(mp_buf *buf, const unsigned char *s, size_t len) {
 }
 
 void mp_buf_free(mp_buf *buf) {
-    free(buf->b);
-    free(buf);
+    mp_realloc(buf->L, buf->b, buf->len, 0); /* realloc to 0 = free */
+    mp_realloc(buf->L, buf, sizeof(*buf), 0);
 }
 
-/* ------------------------------ String cursor ----------------------------------
+/* ---------------------------- String cursor ----------------------------------
  * This simple data structure is used for parsing. Basically you create a cursor
  * using a string pointer and a length, then it is possible to access the
  * current string position with cursor->p, check the remaining length
@@ -102,7 +152,7 @@ void mp_buf_free(mp_buf *buf) {
  * be used to report errors. */
 
 #define MP_CUR_ERROR_NONE   0
-#define MP_CUR_ERROR_EOF    1   /* Not enough data to complete the opereation. */
+#define MP_CUR_ERROR_EOF    1   /* Not enough data to complete operation. */
 #define MP_CUR_ERROR_BADFMT 2   /* Bad data format */
 
 typedef struct mp_cur {
@@ -111,22 +161,15 @@ typedef struct mp_cur {
     int err;
 } mp_cur;
 
-static mp_cur *mp_cur_new(const unsigned char *s, size_t len) {
-    mp_cur *cursor = malloc(sizeof(*cursor));
-
+static void mp_cur_init(mp_cur *cursor, const unsigned char *s, size_t len) {
     cursor->p = s;
     cursor->left = len;
     cursor->err = MP_CUR_ERROR_NONE;
-    return cursor;
-}
-
-static void mp_cur_free(mp_cur *cursor) {
-    free(cursor);
 }
 
 #define mp_cur_consume(_c,_len) do { _c->p += _len; _c->left -= _len; } while(0)
 
-/* When there is not enough room we set an error in the cursor and return, this
+/* When there is not enough room we set an error in the cursor and return. This
  * is very common across the code so we have a macro to make the code look
  * a bit simpler. */
 #define mp_cur_need(_c,_len) do { \
@@ -136,7 +179,7 @@ static void mp_cur_free(mp_cur *cursor) {
     } \
 } while(0)
 
-/* --------------------------- Low level MP encoding -------------------------- */
+/* ------------------------- Low level MP encoding -------------------------- */
 
 static void mp_encode_bytes(mp_buf *buf, const unsigned char *s, size_t len) {
     unsigned char hdr[5];
@@ -299,7 +342,7 @@ static void mp_encode_map(mp_buf *buf, int64_t n) {
     mp_buf_append(buf,b,enclen);
 }
 
-/* ----------------------------- Lua types encoding --------------------------- */
+/* --------------------------- Lua types encoding --------------------------- */
 
 static void mp_encode_lua_string(lua_State *L, mp_buf *buf) {
     size_t len;
@@ -314,13 +357,26 @@ static void mp_encode_lua_bool(lua_State *L, mp_buf *buf) {
     mp_buf_append(buf,&b,1);
 }
 
+/* Lua 5.3 has a built in 64-bit integer type */
+static void mp_encode_lua_integer(lua_State *L, mp_buf *buf) {
+#if (LUA_VERSION_NUM < 503) && BITS_32
+    lua_Number i = lua_tonumber(L,-1);
+#else
+    lua_Integer i = lua_tointeger(L,-1);
+#endif
+    mp_encode_int(buf, (int64_t)i);
+}
+
+/* Lua 5.2 and lower only has 64-bit doubles, so we need to
+ * detect if the double may be representable as an int
+ * for Lua < 5.3 */
 static void mp_encode_lua_number(lua_State *L, mp_buf *buf) {
     lua_Number n = lua_tonumber(L,-1);
 
-    if (floor(n) != n) {
-        mp_encode_double(buf,(double)n);
+    if (IS_INT64_EQUIVALENT(n)) {
+        mp_encode_lua_integer(L, buf);
     } else {
-        mp_encode_int(buf,(int64_t)n);
+        mp_encode_double(buf,(double)n);
     }
 }
 
@@ -328,7 +384,11 @@ static void mp_encode_lua_type(lua_State *L, mp_buf *buf, int level);
 
 /* Convert a lua table into a message pack list. */
 static void mp_encode_lua_table_as_array(lua_State *L, mp_buf *buf, int level) {
+#if LUA_VERSION_NUM < 502
     size_t len = lua_objlen(L,-1), j;
+#else
+    size_t len = lua_rawlen(L,-1), j;
+#endif
 
     mp_encode_array(buf,len);
     for (j = 1; j <= len; j++) {
@@ -345,7 +405,7 @@ static void mp_encode_lua_table_as_map(lua_State *L, mp_buf *buf, int level) {
     /* First step: count keys into table. No other way to do it with the
      * Lua API, we need to iterate a first time. Note that an alternative
      * would be to do a single run, and then hack the buffer to insert the
-     * map opcodes for message pack. Too hachish for this lib. */
+     * map opcodes for message pack. Too hackish for this lib. */
     lua_pushnil(L);
     while(lua_next(L,-2)) {
         lua_pop(L,1); /* remove value, keep key for next iteration. */
@@ -367,30 +427,43 @@ static void mp_encode_lua_table_as_map(lua_State *L, mp_buf *buf, int level) {
  * of keys from numerical keys from 1 up to N, with N being the total number
  * of elements, without any hole in the middle. */
 static int table_is_an_array(lua_State *L) {
-    long count = 0, max = 0, idx = 0;
+    int count = 0, max = 0;
+#if LUA_VERSION_NUM < 503
     lua_Number n;
+#else
+    lua_Integer n;
+#endif
+
+    /* Stack top on function entry */
+    int stacktop;
+
+    stacktop = lua_gettop(L);
 
     lua_pushnil(L);
     while(lua_next(L,-2)) {
         /* Stack: ... key value */
         lua_pop(L,1); /* Stack: ... key */
-        if (lua_type(L,-1) != LUA_TNUMBER) goto not_array;
-        n = lua_tonumber(L,-1);
-        idx = n;
-        if (idx != n || idx < 1) goto not_array;
+        /* The <= 0 check is valid here because we're comparing indexes. */
+#if LUA_VERSION_NUM < 503
+        if ((LUA_TNUMBER != lua_type(L,-1)) || (n = lua_tonumber(L, -1)) <= 0 ||
+            !IS_INT_EQUIVALENT(n))
+#else
+        if (!lua_isinteger(L,-1) || (n = lua_tointeger(L, -1)) <= 0)
+#endif
+        {
+            lua_settop(L, stacktop);
+            return 0;
+        }
+        max = (n > max ? n : max);
         count++;
-        max = idx;
     }
     /* We have the total number of elements in "count". Also we have
-     * the max index encountered in "idx". We can't reach this code
+     * the max index encountered in "max". We can't reach this code
      * if there are indexes <= 0. If you also note that there can not be
-     * repeated keys into a table, you have that if idx==count you are sure
+     * repeated keys into a table, you have that if max==count you are sure
      * that there are all the keys form 1 to count (both included). */
-    return idx == count;
-
-not_array:
-    lua_pop(L,1);
-    return 0;
+    lua_settop(L, stacktop);
+    return max == count;
 }
 
 /* If the length operator returns non-zero, that is, there is at least
@@ -405,6 +478,7 @@ static void mp_encode_lua_table(lua_State *L, mp_buf *buf, int level) {
 
 static void mp_encode_lua_null(lua_State *L, mp_buf *buf) {
     unsigned char b[1];
+    (void)L;
 
     b[0] = 0xc0;
     mp_buf_append(buf,b,1);
@@ -413,29 +487,65 @@ static void mp_encode_lua_null(lua_State *L, mp_buf *buf) {
 static void mp_encode_lua_type(lua_State *L, mp_buf *buf, int level) {
     int t = lua_type(L,-1);
 
-    /* Limit the encoding of nested tables to a specfiied maximum depth, so that
+    /* Limit the encoding of nested tables to a specified maximum depth, so that
      * we survive when called against circular references in tables. */
     if (t == LUA_TTABLE && level == LUACMSGPACK_MAX_NESTING) t = LUA_TNIL;
     switch(t) {
     case LUA_TSTRING: mp_encode_lua_string(L,buf); break;
     case LUA_TBOOLEAN: mp_encode_lua_bool(L,buf); break;
-    case LUA_TNUMBER: mp_encode_lua_number(L,buf); break;
+    case LUA_TNUMBER:
+    #if LUA_VERSION_NUM < 503
+        mp_encode_lua_number(L,buf); break;
+    #else
+        if (lua_isinteger(L, -1)) {
+            mp_encode_lua_integer(L, buf);
+        } else {
+            mp_encode_lua_number(L, buf);
+        }
+        break;
+    #endif
     case LUA_TTABLE: mp_encode_lua_table(L,buf,level); break;
     default: mp_encode_lua_null(L,buf); break;
     }
     lua_pop(L,1);
 }
 
+/*
+ * Packs all arguments as a stream for multiple upacking later.
+ * Returns error if no arguments provided.
+ */
 static int mp_pack(lua_State *L) {
-    mp_buf *buf = mp_buf_new();
+    int nargs = lua_gettop(L);
+    int i;
+    mp_buf *buf;
 
-    mp_encode_lua_type(L,buf,0);
-    lua_pushlstring(L,(char*)buf->b,buf->len);
+    if (nargs == 0)
+        return luaL_argerror(L, 0, "MessagePack pack needs input.");
+
+    buf = mp_buf_new(L);
+    for(i = 1; i <= nargs; i++) {
+        /* Copy argument i to top of stack for _encode processing;
+         * the encode function pops it from the stack when complete. */
+        lua_pushvalue(L, i);
+
+        mp_encode_lua_type(L,buf,0);
+
+        lua_pushlstring(L,(char*)buf->b,buf->len);
+
+        /* Reuse the buffer for the next operation by
+         * setting its free count to the total buffer size
+         * and the current position to zero. */
+        buf->free += buf->len;
+        buf->len = 0;
+    }
     mp_buf_free(buf);
+
+    /* Concatenate all nargs buffers together */
+    lua_concat(L, nargs);
     return 1;
 }
 
-/* --------------------------------- Decoding --------------------------------- */
+/* ------------------------------- Decoding --------------------------------- */
 
 void mp_decode_to_lua_type(lua_State *L, mp_cur *c);
 
@@ -466,34 +576,44 @@ void mp_decode_to_lua_hash(lua_State *L, mp_cur *c, size_t len) {
  * a Lua type, that is left as the only result on the stack. */
 void mp_decode_to_lua_type(lua_State *L, mp_cur *c) {
     mp_cur_need(c,1);
+
+    /* If we return more than 18 elements, we must resize the stack to
+     * fit all our return values.  But, there is no way to
+     * determine how many objects a msgpack will unpack to up front, so
+     * we request a +1 larger stack on each iteration (noop if stack is
+     * big enough, and when stack does require resize it doubles in size) */
+    luaL_checkstack(L, 1,
+        "too many return values at once; "
+        "use unpack_one or unpack_limit instead.");
+
     switch(c->p[0]) {
     case 0xcc:  /* uint 8 */
         mp_cur_need(c,2);
-        lua_pushnumber(L,c->p[1]);
+        lua_pushunsigned(L,c->p[1]);
         mp_cur_consume(c,2);
         break;
     case 0xd0:  /* int 8 */
         mp_cur_need(c,2);
-        lua_pushnumber(L,(char)c->p[1]);
+        lua_pushinteger(L,(char)c->p[1]);
         mp_cur_consume(c,2);
         break;
     case 0xcd:  /* uint 16 */
         mp_cur_need(c,3);
-        lua_pushnumber(L,
+        lua_pushunsigned(L,
             (c->p[1] << 8) |
              c->p[2]);
         mp_cur_consume(c,3);
         break;
     case 0xd1:  /* int 16 */
         mp_cur_need(c,3);
-        lua_pushnumber(L,(int16_t)
+        lua_pushinteger(L,(int16_t)
             (c->p[1] << 8) |
              c->p[2]);
         mp_cur_consume(c,3);
         break;
     case 0xce:  /* uint 32 */
         mp_cur_need(c,5);
-        lua_pushnumber(L,
+        lua_pushunsigned(L,
             ((uint32_t)c->p[1] << 24) |
             ((uint32_t)c->p[2] << 16) |
             ((uint32_t)c->p[3] << 8) |
@@ -502,7 +622,7 @@ void mp_decode_to_lua_type(lua_State *L, mp_cur *c) {
         break;
     case 0xd2:  /* int 32 */
         mp_cur_need(c,5);
-        lua_pushnumber(L,
+        lua_pushinteger(L,
             ((int32_t)c->p[1] << 24) |
             ((int32_t)c->p[2] << 16) |
             ((int32_t)c->p[3] << 8) |
@@ -511,7 +631,7 @@ void mp_decode_to_lua_type(lua_State *L, mp_cur *c) {
         break;
     case 0xcf:  /* uint 64 */
         mp_cur_need(c,9);
-        lua_pushnumber(L,
+        lua_pushunsigned(L,
             ((uint64_t)c->p[1] << 56) |
             ((uint64_t)c->p[2] << 48) |
             ((uint64_t)c->p[3] << 40) |
@@ -524,7 +644,11 @@ void mp_decode_to_lua_type(lua_State *L, mp_cur *c) {
         break;
     case 0xd3:  /* int 64 */
         mp_cur_need(c,9);
+#if LUA_VERSION_NUM < 503
         lua_pushnumber(L,
+#else
+        lua_pushinteger(L,
+#endif
             ((int64_t)c->p[1] << 56) |
             ((int64_t)c->p[2] << 48) |
             ((int64_t)c->p[3] << 40) |
@@ -630,10 +754,10 @@ void mp_decode_to_lua_type(lua_State *L, mp_cur *c) {
         break;
     default:    /* types that can't be idenitified by first byte value. */
         if ((c->p[0] & 0x80) == 0) {   /* positive fixnum */
-            lua_pushnumber(L,c->p[0]);
+            lua_pushunsigned(L,c->p[0]);
             mp_cur_consume(c,1);
         } else if ((c->p[0] & 0xe0) == 0xe0) {  /* negative fixnum */
-            lua_pushnumber(L,(signed char)c->p[0]);
+            lua_pushinteger(L,(signed char)c->p[0]);
             mp_cur_consume(c,1);
         } else if ((c->p[0] & 0xe0) == 0xa0) {  /* fix raw */
             size_t l = c->p[0] & 0x1f;
@@ -654,54 +778,163 @@ void mp_decode_to_lua_type(lua_State *L, mp_cur *c) {
     }
 }
 
-static int mp_unpack(lua_State *L) {
+static int mp_unpack_full(lua_State *L, int limit, int offset) {
     size_t len;
-    const unsigned char *s;
-    mp_cur *c;
+    const char *s;
+    mp_cur c;
+    int cnt; /* Number of objects unpacked */
+    int decode_all = (!limit && !offset);
 
-    if (!lua_isstring(L,-1)) {
-        lua_pushstring(L,"MessagePack decoding needs a string as input.");
-        lua_error(L);
+    s = luaL_checklstring(L,1,&len); /* if no match, exits */
+
+    if (offset < 0 || limit < 0) /* requesting negative off or lim is invalid */
+        return luaL_error(L,
+            "Invalid request to unpack with offset of %d and limit of %d.",
+            offset, len);
+    else if (offset > len)
+        return luaL_error(L,
+            "Start offset %d greater than input length %d.", offset, len);
+
+    if (decode_all) limit = INT_MAX;
+
+    mp_cur_init(&c,(const unsigned char *)s+offset,len-offset);
+
+    /* We loop over the decode because this could be a stream
+     * of multiple top-level values serialized together */
+    for(cnt = 0; c.left > 0 && cnt < limit; cnt++) {
+        mp_decode_to_lua_type(L,&c);
+
+        if (c.err == MP_CUR_ERROR_EOF) {
+            return luaL_error(L,"Missing bytes in input.");
+        } else if (c.err == MP_CUR_ERROR_BADFMT) {
+            return luaL_error(L,"Bad data format in input.");
+        }
     }
 
-    s = (const unsigned char*) lua_tolstring(L,-1,&len);
-    c = mp_cur_new(s,len);
-    mp_decode_to_lua_type(L,c);
-    
-    if (c->err == MP_CUR_ERROR_EOF) {
-        mp_cur_free(c);
-        lua_pushstring(L,"Missing bytes in input.");
-        lua_error(L);
-    } else if (c->err == MP_CUR_ERROR_BADFMT) {
-        mp_cur_free(c);
-        lua_pushstring(L,"Bad data format in input.");
-        lua_error(L);
-    } else if (c->left != 0) {
-        mp_cur_free(c);
-        lua_pushstring(L,"Extra bytes in input.");
-        lua_error(L);
+    if (!decode_all) {
+        /* c->left is the remaining size of the input buffer.
+         * subtract the entire buffer size from the unprocessed size
+         * to get our next start offset */
+        int offset = len - c.left;
+        /* Return offset -1 when we have have processed the entire buffer. */
+        lua_pushinteger(L, c.left == 0 ? -1 : offset);
+        /* Results are returned with the arg elements still
+         * in place. Lua takes care of only returning
+         * elements above the args for us.
+         * In this case, we have one arg on the stack
+         * for this function, so we insert our first return
+         * value at position 2. */
+        lua_insert(L, 2);
+        cnt += 1; /* increase return count by one to make room for offset */
     }
-    mp_cur_free(c);
-    return 1;
+
+    return cnt;
 }
 
-/* ---------------------------------------------------------------------------- */
+static int mp_unpack(lua_State *L) {
+    return mp_unpack_full(L, 0, 0);
+}
 
-static const struct luaL_reg thislib[] = {
+static int mp_unpack_one(lua_State *L) {
+    int offset = luaL_optint(L, 2, 0);
+    /* Variable pop because offset may not exist */
+    lua_pop(L, lua_gettop(L)-1);
+    return mp_unpack_full(L, 1, offset);
+}
+
+static int mp_unpack_limit(lua_State *L) {
+    int limit = luaL_checkint(L, 2);
+    int offset = luaL_optint(L, 3, 0);
+    /* Variable pop because offset may not exist */
+    lua_pop(L, lua_gettop(L)-1);
+
+    return mp_unpack_full(L, limit, offset);
+}
+
+static int mp_safe(lua_State *L) {
+    int argc, err, total_results;
+
+    argc = lua_gettop(L);
+
+    /* This adds our function to the bottom of the stack
+     * (the "call this function" position) */
+    lua_pushvalue(L, lua_upvalueindex(1));
+    lua_insert(L, 1);
+
+    err = lua_pcall(L, argc, LUA_MULTRET, 0);
+    total_results = lua_gettop(L);
+
+    if (!err) {
+        return total_results;
+    } else {
+        lua_pushnil(L);
+        lua_insert(L,-2);
+        return 2;
+    }
+}
+
+/* -------------------------------------------------------------------------- */
+static const struct luaL_Reg cmds[] = {
     {"pack", mp_pack},
     {"unpack", mp_unpack},
-    {NULL, NULL}
+    {"unpack_one", mp_unpack_one},
+    {"unpack_limit", mp_unpack_limit},
+    {0}
 };
 
-LUALIB_API int luaopen_cmsgpack (lua_State *L) {
-    luaL_register(L, "cmsgpack", thislib);
+static int luaopen_create(lua_State *L) {
+    int i;
+    /* Manually construct our module table instead of
+     * relying on _register or _newlib */
+    lua_newtable(L);
 
+    for (i = 0; i < (sizeof(cmds)/sizeof(*cmds) - 1); i++) {
+        lua_pushcfunction(L, cmds[i].func);
+        lua_setfield(L, -2, cmds[i].name);
+    }
+
+    /* Add metadata */
+    lua_pushliteral(L, LUACMSGPACK_NAME);
+    lua_setfield(L, -2, "_NAME");
     lua_pushliteral(L, LUACMSGPACK_VERSION);
     lua_setfield(L, -2, "_VERSION");
     lua_pushliteral(L, LUACMSGPACK_COPYRIGHT);
     lua_setfield(L, -2, "_COPYRIGHT");
     lua_pushliteral(L, LUACMSGPACK_DESCRIPTION);
-    lua_setfield(L, -2, "_DESCRIPTION"); 
+    lua_setfield(L, -2, "_DESCRIPTION");
+    return 1;
+}
+
+LUALIB_API int luaopen_cmsgpack(lua_State *L) {
+    luaopen_create(L);
+
+#if LUA_VERSION_NUM < 502
+    /* Register name globally for 5.1 */
+    lua_pushvalue(L, -1);
+    lua_setglobal(L, LUACMSGPACK_NAME);
+#endif
+
+    return 1;
+}
+
+LUALIB_API int luaopen_cmsgpack_safe(lua_State *L) {
+    int i;
+
+    luaopen_cmsgpack(L);
+
+    /* Wrap all functions in the safe handler */
+    for (i = 0; i < (sizeof(cmds)/sizeof(*cmds) - 1); i++) {
+        lua_getfield(L, -1, cmds[i].name);
+        lua_pushcclosure(L, mp_safe, 1);
+        lua_setfield(L, -2, cmds[i].name);
+    }
+
+#if LUA_VERSION_NUM < 502
+    /* Register name globally for 5.1 */
+    lua_pushvalue(L, -1);
+    lua_setglobal(L, LUACMSGPACK_SAFE_NAME);
+#endif
+
     return 1;
 }
 

deps/lua/src/strbuf.c

@@ -1,6 +1,6 @@
-/* strbuf - string buffer routines
+/* strbuf - String buffer routines
  *
- * Copyright (c) 2010-2011  Mark Pulford <mark@kyne.com.au>
+ * Copyright (c) 2010-2012  Mark Pulford <mark@kyne.com.au>
  *
  * Permission is hereby granted, free of charge, to any person obtaining
  * a copy of this software and associated documentation files (the
@@ -29,7 +29,7 @@
 
 #include "strbuf.h"
 
-void die(const char *fmt, ...)
+static void die(const char *fmt, ...)
 {
     va_list arg;
 

deps/lua/src/strbuf.h

@@ -1,6 +1,6 @@
 /* strbuf - String buffer routines
  *
- * Copyright (c) 2010-2011  Mark Pulford <mark@kyne.com.au>
+ * Copyright (c) 2010-2012  Mark Pulford <mark@kyne.com.au>
  *
  * Permission is hereby granted, free of charge, to any person obtaining
  * a copy of this software and associated documentation files (the
@@ -62,7 +62,9 @@ extern void strbuf_resize(strbuf_t *s, int len);
 static int strbuf_empty_length(strbuf_t *s);
 static int strbuf_length(strbuf_t *s);
 static char *strbuf_string(strbuf_t *s, int *len);
-static  void strbuf_ensure_empty_length(strbuf_t *s, int len);
+static void strbuf_ensure_empty_length(strbuf_t *s, int len);
+static char *strbuf_empty_ptr(strbuf_t *s);
+static void strbuf_extend_length(strbuf_t *s, int len);
 
 /* Update */
 extern void strbuf_append_fmt(strbuf_t *s, int len, const char *fmt, ...);
@@ -96,6 +98,16 @@ static inline void strbuf_ensure_empty_length(strbuf_t *s, int len)
         strbuf_resize(s, s->length + len);
 }
 
+static inline char *strbuf_empty_ptr(strbuf_t *s)
+{
+    return s->buf + s->length;
+}
+
+static inline void strbuf_extend_length(strbuf_t *s, int len)
+{
+    s->length += len;
+}
+
 static inline int strbuf_length(strbuf_t *s)
 {
     return s->length;

redis.conf

@@ -15,7 +15,7 @@
 ################################## INCLUDES ###################################
 
 # Include one or more other config files here.  This is useful if you
-# have a standard template that goes to all Redis server but also need
+# have a standard template that goes to all Redis servers but also need
 # to customize a few per-server settings.  Include files can include
 # other files, so use this wisely.
 #
@@ -68,7 +68,7 @@ tcp-backlog 511
 # on a unix socket when not specified.
 #
 # unixsocket /tmp/redis.sock
-# unixsocketperm 755
+# unixsocketperm 700
 
 # Close the connection after a client is idle for N seconds (0 to disable)
 timeout 0
@@ -131,7 +131,7 @@ databases 16
 #   after 300 sec (5 min) if at least 10 keys changed
 #   after 60 sec if at least 10000 keys changed
 #
-#   Note: you can disable saving at all commenting all the "save" lines.
+#   Note: you can disable saving completely by commenting out all "save" lines.
 #
 #   It is also possible to remove all the previously configured save
 #   points by adding a save directive with a single empty string argument
@@ -180,18 +180,27 @@ dbfilename dump.rdb
 #
 # The DB will be written inside this directory, with the filename specified
 # above using the 'dbfilename' configuration directive.
-# 
+#
 # The Append Only File will also be created inside this directory.
-# 
+#
 # Note that you must specify a directory here, not a file name.
 dir ./
 
 ################################# REPLICATION #################################
 
 # Master-Slave replication. Use slaveof to make a Redis instance a copy of
-# another Redis server. Note that the configuration is local to the slave
-# so for example it is possible to configure the slave to save the DB with a
-# different interval, or to listen to another port, and so on.
+# another Redis server. A few things to understand ASAP about Redis replication.
+#
+# 1) Redis replication is asynchronous, but you can configure a master to
+#    stop accepting writes if it appears to be not connected with at least
+#    a given number of slaves.
+# 2) Redis slaves are able to perform a partial resynchronization with the
+#    master if the replication link is lost for a relatively small amount of
+#    time. You may want to configure the replication backlog size (see the next
+#    sections of this file) with a sensible value depending on your needs.
+# 3) Replication is automatic and does not need user intervention. After a
+#    network partition slaves automatically try to reconnect to masters
+#    and resynchronize with them.
 #
 # slaveof <masterip> <masterport>
 
@@ -231,6 +240,49 @@ slave-serve-stale-data yes
 # administrative / dangerous commands.
 slave-read-only yes
 
+# Replication SYNC strategy: disk or socket.
+#
+# -------------------------------------------------------
+# WARNING: DISKLESS REPLICATION IS EXPERIMENTAL CURRENTLY
+# -------------------------------------------------------
+#
+# New slaves and reconnecting slaves that are not able to continue the replication
+# process just receiving differences, need to do what is called a "full
+# synchronization". An RDB file is transmitted from the master to the slaves.
+# The transmission can happen in two different ways:
+#
+# 1) Disk-backed: The Redis master creates a new process that writes the RDB
+#                 file on disk. Later the file is transferred by the parent
+#                 process to the slaves incrementally.
+# 2) Diskless: The Redis master creates a new process that directly writes the
+#              RDB file to slave sockets, without touching the disk at all.
+#
+# With disk-backed replication, while the RDB file is generated, more slaves
+# can be queued and served with the RDB file as soon as the current child producing
+# the RDB file finishes its work. With diskless replication instead once
+# the transfer starts, new slaves arriving will be queued and a new transfer
+# will start when the current one terminates.
+#
+# When diskless replication is used, the master waits a configurable amount of
+# time (in seconds) before starting the transfer in the hope that multiple slaves
+# will arrive and the transfer can be parallelized.
+#
+# With slow disks and fast (large bandwidth) networks, diskless replication
+# works better.
+repl-diskless-sync no
+
+# When diskless replication is enabled, it is possible to configure the delay
+# the server waits in order to spawn the child that trnasfers the RDB via socket
+# to the slaves.
+#
+# This is important since once the transfer starts, it is not possible to serve
+# new slaves arriving, that will be queued for the next RDB transfer, so the server
+# waits a delay in order to let more slaves arrive.
+#
+# The delay is specified in seconds, and by default is 5 seconds. To disable
+# it entirely just set it to 0 seconds and the transfer will start ASAP.
+repl-diskless-sync-delay 5
+
 # Slaves send PINGs to server in a predefined interval. It's possible to change
 # this interval with the repl_ping_slave_period option. The default value is 10
 # seconds.
@@ -270,7 +322,7 @@ repl-disable-tcp-nodelay no
 # resync is enough, just passing the portion of data the slave missed while
 # disconnected.
 #
-# The biggest the replication backlog, the longer the time the slave can be
+# The bigger the replication backlog, the longer the time the slave can be
 # disconnected and later be able to perform a partial resynchronization.
 #
 # The backlog is only allocated once there is at least a slave connected.
@@ -309,7 +361,7 @@ slave-priority 100
 # The lag in seconds, that must be <= the specified value, is calculated from
 # the last ping received from the slave, that is usually sent every second.
 #
-# This option does not GUARANTEES that N replicas will accept the write, but
+# This option does not GUARANTEE that N replicas will accept the write, but
 # will limit the window of exposure for lost writes in case not enough slaves
 # are available, to the specified number of seconds.
 #
@@ -331,7 +383,7 @@ slave-priority 100
 #
 # This should stay commented out for backward compatibility and because most
 # people do not need auth (e.g. they run their own servers).
-# 
+#
 # Warning: since Redis is pretty fast an outside user can try up to
 # 150k passwords per second against a good box. This means that you should
 # use a very strong password otherwise it will be very easy to break.
@@ -397,18 +449,18 @@ slave-priority 100
 
 # MAXMEMORY POLICY: how Redis will select what to remove when maxmemory
 # is reached. You can select among five behaviors:
-# 
+#
 # volatile-lru -> remove the key with an expire set using an LRU algorithm
-# allkeys-lru -> remove any key accordingly to the LRU algorithm
+# allkeys-lru -> remove any key according to the LRU algorithm
 # volatile-random -> remove a random key with an expire set
 # allkeys-random -> remove a random key, any key
 # volatile-ttl -> remove the key with the nearest expire time (minor TTL)
 # noeviction -> don't expire at all, just return an error on write operations
-# 
-# Note: with any of the above policies, Redis will return an error on write
-#       operations, when there are not suitable keys for eviction.
 #
-#       At the date of writing this commands are: set setnx setex append
+# Note: with any of the above policies, Redis will return an error on write
+#       operations, when there are no suitable keys for eviction.
+#
+#       At the date of writing these commands are: set setnx setex append
 #       incr decr rpush lpush rpushx lpushx linsert lset rpoplpush sadd
 #       sinter sinterstore sunion sunionstore sdiff sdiffstore zadd zincrby
 #       zunionstore zinterstore hset hsetnx hmset hincrby incrby decrby
@@ -456,13 +508,13 @@ appendonly no
 appendfilename "appendonly.aof"
 
 # The fsync() call tells the Operating System to actually write data on disk
-# instead to wait for more data in the output buffer. Some OS will really flush 
+# instead of waiting for more data in the output buffer. Some OS will really flush
 # data on disk, some other OS will just try to do it ASAP.
 #
 # Redis supports three different modes:
 #
 # no: don't fsync, just let the OS flush the data when it wants. Faster.
-# always: fsync after every write to the append only log . Slow, Safest.
+# always: fsync after every write to the append only log. Slow, Safest.
 # everysec: fsync only one time every second. Compromise.
 #
 # The default is "everysec", as that's usually the right compromise between
@@ -497,7 +549,7 @@ appendfsync everysec
 # the same as "appendfsync none". In practical terms, this means that it is
 # possible to lose up to 30 seconds of log in the worst scenario (with the
 # default Linux settings).
-# 
+#
 # If you have latency problems turn this to "yes". Otherwise leave it as
 # "no" that is the safest pick from the point of view of durability.
 
@@ -506,7 +558,7 @@ no-appendfsync-on-rewrite no
 # Automatic rewrite of the append only file.
 # Redis is able to automatically rewrite the log file implicitly calling
 # BGREWRITEAOF when the AOF log size grows by the specified percentage.
-# 
+#
 # This is how it works: Redis remembers the size of the AOF file after the
 # latest rewrite (if no rewrite has happened since the restart, the size of
 # the AOF at startup is used).
@@ -523,6 +575,30 @@ no-appendfsync-on-rewrite no
 auto-aof-rewrite-percentage 100
 auto-aof-rewrite-min-size 64mb
 
+# An AOF file may be found to be truncated at the end during the Redis
+# startup process, when the AOF data gets loaded back into memory.
+# This may happen when the system where Redis is running
+# crashes, especially when an ext4 filesystem is mounted without the
+# data=ordered option (however this can't happen when Redis itself
+# crashes or aborts but the operating system still works correctly).
+#
+# Redis can either exit with an error when this happens, or load as much
+# data as possible (the default now) and start if the AOF file is found
+# to be truncated at the end. The following option controls this behavior.
+#
+# If aof-load-truncated is set to yes, a truncated AOF file is loaded and
+# the Redis server starts emitting a log to inform the user of the event.
+# Otherwise if the option is set to no, the server aborts with an error
+# and refuses to start. When the option is set to no, the user requires
+# to fix the AOF file using the "redis-check-aof" utility before to restart
+# the server.
+#
+# Note that if the AOF file will be found to be corrupted in the middle
+# the server will still exit with an error. This option only applies when
+# Redis will try to read more data from the AOF file but not enough bytes
+# will be found.
+aof-load-truncated yes
+
 ################################ LUA SCRIPTING  ###############################
 
 # Max execution time of a Lua script in milliseconds.
@@ -531,11 +607,11 @@ auto-aof-rewrite-min-size 64mb
 # still in execution after the maximum allowed time and will start to
 # reply to queries with an error.
 #
-# When a long running script exceed the maximum execution time only the
+# When a long running script exceeds the maximum execution time only the
 # SCRIPT KILL and SHUTDOWN NOSAVE commands are available. The first can be
 # used to stop a script that did not yet called write commands. The second
-# is the only way to shut down the server in the case a write commands was
-# already issue by the script but the user don't want to wait for the natural
+# is the only way to shut down the server in the case a write command was
+# already issued by the script but the user doesn't want to wait for the natural
 # termination of the script.
 #
 # Set it to 0 or a negative value for unlimited execution without warnings.
@@ -552,12 +628,12 @@ lua-time-limit 5000
 # Every cluster node has a cluster configuration file. This file is not
 # intended to be edited by hand. It is created and updated by Redis nodes.
 # Every Redis Cluster node requires a different cluster configuration file.
-# Make sure that instances running in the same system does not have
+# Make sure that instances running in the same system do not have
 # overlapping cluster configuration file names.
 #
 # cluster-config-file nodes-6379.conf
 
-# Cluster node timeout is the amount of milliseconds a node must be unreachable 
+# Cluster node timeout is the amount of milliseconds a node must be unreachable
 # for it to be considered in failure state.
 # Most other internal time limits are multiple of the node timeout.
 #
@@ -627,6 +703,19 @@ lua-time-limit 5000
 #
 # cluster-migration-barrier 1
 
+# By default Redis Cluster nodes stop accepting queries if they detect there
+# is at least an hash slot uncovered (no available node is serving it).
+# This way if the cluster is partially down (for example a range of hash slots
+# are no longer covered) all the cluster becomes, eventually, unavailable.
+# It automatically returns available as soon as all the slots are covered again.
+#
+# However sometimes you want the subset of the cluster which is working,
+# to continue to accept queries for the part of the key space that is still
+# covered. In order to do so, just set the cluster-require-full-coverage
+# option to no.
+#
+# cluster-require-full-coverage yes
+
 # In order to setup your cluster make sure to read the documentation
 # available at http://redis.io web site.
 
@@ -638,7 +727,7 @@ lua-time-limit 5000
 # but just the time needed to actually execute the command (this is the only
 # stage of command execution where the thread is blocked and can not serve
 # other requests in the meantime).
-# 
+#
 # You can configure the slow log with two parameters: one tells Redis
 # what is the execution time, in microseconds, to exceed in order for the
 # command to get logged, and the other parameter is the length of the
@@ -678,8 +767,8 @@ latency-monitor-threshold 0
 ############################# Event notification ##############################
 
 # Redis can notify Pub/Sub clients about events happening in the key space.
-# This feature is documented at http://redis.io/topics/keyspace-events
-# 
+# This feature is documented at http://redis.io/topics/notifications
+#
 # For instance if keyspace events notification is enabled, and a client
 # performs a DEL operation on key "foo" stored in the Database 0, two
 # messages will be published via Pub/Sub:
@@ -703,8 +792,8 @@ latency-monitor-threshold 0
 #  A     Alias for g$lshzxe, so that the "AKE" string means all the events.
 #
 #  The "notify-keyspace-events" takes as argument a string that is composed
-#  by zero or multiple characters. The empty string means that notifications
-#  are disabled at all.
+#  of zero or multiple characters. The empty string means that notifications
+#  are disabled.
 #
 #  Example: to enable list and generic events, from the point of view of the
 #           event name, use:
@@ -736,7 +825,7 @@ list-max-ziplist-entries 512
 list-max-ziplist-value 64
 
 # Sets have a special encoding in just one case: when a set is composed
-# of just strings that happens to be integers in radix 10 in the range
+# of just strings that happen to be integers in radix 10 in the range
 # of 64 bit signed integers.
 # The following configuration setting sets the limit in the size of the
 # set in order to use this special memory saving encoding.
@@ -754,7 +843,7 @@ zset-max-ziplist-value 64
 #
 # A value greater than 16000 is totally useless, since at that point the
 # dense representation is more memory efficient.
-# 
+#
 # The suggested value is ~ 3000 in order to have the benefits of
 # the space efficient encoding without slowing down too much PFADD,
 # which is O(N) with the sparse encoding. The value can be raised to
@@ -769,13 +858,13 @@ hll-sparse-max-bytes 3000
 # that is rehashing, the more rehashing "steps" are performed, so if the
 # server is idle the rehashing is never complete and some more memory is used
 # by the hash table.
-# 
+#
 # The default is to use this millisecond 10 times every second in order to
-# active rehashing the main dictionaries, freeing memory when possible.
+# actively rehash the main dictionaries, freeing memory when possible.
 #
 # If unsure:
 # use "activerehashing no" if you have hard latency requirements and it is
-# not a good thing in your environment that Redis can reply form time to time
+# not a good thing in your environment that Redis can reply from time to time
 # to queries with 2 milliseconds delay.
 #
 # use "activerehashing yes" if you don't have such hard requirements but
@@ -824,7 +913,7 @@ client-output-buffer-limit pubsub 32mb 8mb 60
 # never requested, and so forth.
 #
 # Not all tasks are performed with the same frequency, but Redis checks for
-# tasks to perform accordingly to the specified "hz" value.
+# tasks to perform according to the specified "hz" value.
 #
 # By default "hz" is set to 10. Raising the value will use more CPU when
 # Redis is idle, but at the same time will make Redis more responsive when
@@ -841,4 +930,3 @@ hz 10
 # in order to commit the file to the disk more incrementally and avoid
 # big latency spikes.
 aof-rewrite-incremental-fsync yes
-

sentinel.conf

@@ -4,10 +4,32 @@
 # The port that this sentinel instance will run on
 port 26379
 
+# sentinel announce-ip <ip>
+# sentinel announce-port <port>
+#
+# The above two configuration directives are useful in environments where,
+# because of NAT, Sentinel is reachable from outside via a non-local address.
+#
+# When announce-ip is provided, the Sentinel will claim the specified IP address
+# in HELLO messages used to gossip its presence, instead of auto-detecting the
+# local address as it usually does.
+#
+# Similarly when announce-port is provided and is valid and non-zero, Sentinel
+# will announce the specified TCP port.
+#
+# The two options don't need to be used together, if only announce-ip is
+# provided, the Sentinel will announce the specified IP and the server port
+# as specified by the "port" option. If only announce-port is provided, the
+# Sentinel will announce the auto-detected local IP and the specified port.
+#
+# Example:
+#
+# sentinel announce-ip 1.2.3.4
+
 # dir <working-directory>
 # Every long running process should have a well-defined working directory.
 # For Redis Sentinel to chdir to /tmp at startup is the simplest thing
-# for the process to don't interferer with administrative tasks such as
+# for the process to don't interfere with administrative tasks such as
 # unmounting filesystems.
 dir /tmp
 

src/Makefile

@@ -19,7 +19,7 @@ DEPENDENCY_TARGETS=hiredis linenoise lua
 
 # Default settings
 STD=-std=c99 -pedantic
-WARN=-Wall
+WARN=-Wall -W
 OPT=$(OPTIMIZATION)
 
 PREFIX?=/usr/local
@@ -58,17 +58,23 @@ ifeq ($(uname_S),SunOS)
 	# SunOS
 	INSTALL=cp -pf
 	FINAL_CFLAGS+= -D__EXTENSIONS__ -D_XPG6
-	FINAL_LIBS+= -ldl -lnsl -lsocket -lpthread
+	FINAL_LIBS+= -ldl -lnsl -lsocket -lresolv -lpthread
 else
 ifeq ($(uname_S),Darwin)
 	# Darwin (nothing to do)
+else
+ifeq ($(uname_S),AIX)
+        # AIX
+        FINAL_LDFLAGS+= -Wl,-bexpall
+        FINAL_LIBS+= -pthread -lcrypt -lbsd
+
 else
 	# All the other OSes (notably Linux)
 	FINAL_LDFLAGS+= -rdynamic
 	FINAL_LIBS+= -pthread
 endif
 endif
-
+endif
 # Include paths to dependencies
 FINAL_CFLAGS+= -I../deps/hiredis -I../deps/linenoise -I../deps/lua/src
 
@@ -119,7 +125,7 @@ REDIS_CHECK_AOF_OBJ=redis-check-aof.o
 
 all: $(REDIS_SERVER_NAME) $(REDIS_SENTINEL_NAME) $(REDIS_CLI_NAME) $(REDIS_BENCHMARK_NAME) $(REDIS_CHECK_DUMP_NAME) $(REDIS_CHECK_AOF_NAME)
 	@echo ""
-	@echo "Hint: To run 'make test' is a good idea ;)"
+	@echo "Hint: It's a good idea to run 'make test' ;)"
 	@echo ""
 
 .PHONY: all

src/adlist.c

@@ -71,7 +71,7 @@ void listRelease(list *list)
     zfree(list);
 }
 
-/* Add a new node to the list, to head, contaning the specified 'value'
+/* Add a new node to the list, to head, containing the specified 'value'
  * pointer as value.
  *
  * On error, NULL is returned and no operation is performed (i.e. the

src/ae.c

@@ -156,8 +156,9 @@ void aeDeleteFileEvent(aeEventLoop *eventLoop, int fd, int mask)
 {
     if (fd >= eventLoop->setsize) return;
     aeFileEvent *fe = &eventLoop->events[fd];
-
     if (fe->mask == AE_NONE) return;
+
+    aeApiDelEvent(eventLoop, fd, mask);
     fe->mask = fe->mask & (~mask);
     if (fd == eventLoop->maxfd && fe->mask == AE_NONE) {
         /* Update the max fd */
@@ -167,7 +168,6 @@ void aeDeleteFileEvent(aeEventLoop *eventLoop, int fd, int mask)
             if (eventLoop->events[j].mask != AE_NONE) break;
         eventLoop->maxfd = j;
     }
-    aeApiDelEvent(eventLoop, fd, mask);
 }
 
 int aeGetFileEvents(aeEventLoop *eventLoop, int fd) {

src/anet.c

@@ -34,6 +34,7 @@
 #include <sys/socket.h>
 #include <sys/stat.h>
 #include <sys/un.h>
+#include <sys/time.h>
 #include <netinet/in.h>
 #include <netinet/tcp.h>
 #include <arpa/inet.h>
@@ -57,24 +58,37 @@ static void anetSetError(char *err, const char *fmt, ...)
     va_end(ap);
 }
 
-int anetNonBlock(char *err, int fd)
-{
+int anetSetBlock(char *err, int fd, int non_block) {
     int flags;
 
-    /* Set the socket non-blocking.
+    /* Set the socket blocking (if non_block is zero) or non-blocking.
      * Note that fcntl(2) for F_GETFL and F_SETFL can't be
      * interrupted by a signal. */
     if ((flags = fcntl(fd, F_GETFL)) == -1) {
         anetSetError(err, "fcntl(F_GETFL): %s", strerror(errno));
         return ANET_ERR;
     }
-    if (fcntl(fd, F_SETFL, flags | O_NONBLOCK) == -1) {
+
+    if (non_block)
+        flags |= O_NONBLOCK;
+    else
+        flags &= ~O_NONBLOCK;
+
+    if (fcntl(fd, F_SETFL, flags) == -1) {
         anetSetError(err, "fcntl(F_SETFL,O_NONBLOCK): %s", strerror(errno));
         return ANET_ERR;
     }
     return ANET_OK;
 }
 
+int anetNonBlock(char *err, int fd) {
+    return anetSetBlock(err,fd,1);
+}
+
+int anetBlock(char *err, int fd) {
+    return anetSetBlock(err,fd,0);
+}
+
 /* Set TCP keep alive option to detect dead peers. The interval option
  * is only used for Linux as we are using Linux-specific APIs to set
  * the probe send time, interval, and count. */
@@ -117,6 +131,8 @@ int anetKeepAlive(char *err, int fd, int interval)
         anetSetError(err, "setsockopt TCP_KEEPCNT: %s\n", strerror(errno));
         return ANET_ERR;
     }
+#else
+    ((void) interval); /* Avoid unused var warning for non Linux systems. */
 #endif
 
     return ANET_OK;
@@ -163,6 +179,20 @@ int anetTcpKeepAlive(char *err, int fd)
     return ANET_OK;
 }
 
+/* Set the socket send timeout (SO_SNDTIMEO socket option) to the specified
+ * number of milliseconds, or disable it if the 'ms' argument is zero. */
+int anetSendTimeout(char *err, int fd, long long ms) {
+    struct timeval tv;
+
+    tv.tv_sec = ms/1000;
+    tv.tv_usec = (ms%1000)*1000;
+    if (setsockopt(fd, SOL_SOCKET, SO_SNDTIMEO, &tv, sizeof(tv)) == -1) {
+        anetSetError(err, "setsockopt SO_SNDTIMEO: %s", strerror(errno));
+        return ANET_ERR;
+    }
+    return ANET_OK;
+}
+
 /* anetGenericResolve() is called by anetResolve() and anetResolveIP() to
  * do the actual work. It resolves the hostname "host" and set the string
  * representation of the IP address into the buffer pointed by "ipbuf".
@@ -262,7 +292,8 @@ static int anetTcpGenericConnect(char *err, char *addr, int port,
         if (source_addr) {
             int bound = 0;
             /* Using getaddrinfo saves us from self-determining IPv4 vs IPv6 */
-            if ((rv = getaddrinfo(source_addr, NULL, &hints, &bservinfo)) != 0) {
+            if ((rv = getaddrinfo(source_addr, NULL, &hints, &bservinfo)) != 0)
+            {
                 anetSetError(err, "%s", gai_strerror(rv));
                 goto end;
             }
@@ -272,6 +303,7 @@ static int anetTcpGenericConnect(char *err, char *addr, int port,
                     break;
                 }
             }
+            freeaddrinfo(bservinfo);
             if (!bound) {
                 anetSetError(err, "bind: %s", strerror(errno));
                 goto end;
@@ -525,22 +557,36 @@ int anetPeerToString(int fd, char *ip, size_t ip_len, int *port) {
     struct sockaddr_storage sa;
     socklen_t salen = sizeof(sa);
 
-    if (getpeername(fd,(struct sockaddr*)&sa,&salen) == -1) {
-        if (port) *port = 0;
-        ip[0] = '?';
-        ip[1] = '\0';
-        return -1;
-    }
+    if (getpeername(fd,(struct sockaddr*)&sa,&salen) == -1) goto error;
+    if (ip_len == 0) goto error;
+
     if (sa.ss_family == AF_INET) {
         struct sockaddr_in *s = (struct sockaddr_in *)&sa;
         if (ip) inet_ntop(AF_INET,(void*)&(s->sin_addr),ip,ip_len);
         if (port) *port = ntohs(s->sin_port);
-    } else {
+    } else if (sa.ss_family == AF_INET6) {
         struct sockaddr_in6 *s = (struct sockaddr_in6 *)&sa;
         if (ip) inet_ntop(AF_INET6,(void*)&(s->sin6_addr),ip,ip_len);
         if (port) *port = ntohs(s->sin6_port);
+    } else if (sa.ss_family == AF_UNIX) {
+        if (ip) strncpy(ip,"/unixsocket",ip_len);
+        if (port) *port = 0;
+    } else {
+        goto error;
     }
     return 0;
+
+error:
+    if (ip) {
+        if (ip_len >= 2) {
+            ip[0] = '?';
+            ip[1] = '\0';
+        } else if (ip_len == 1) {
+            ip[0] = '\0';
+        }
+    }
+    if (port) *port = 0;
+    return -1;
 }
 
 int anetSockName(int fd, char *ip, size_t ip_len, int *port) {

src/anet.h

@@ -39,10 +39,14 @@
 #define ANET_NONE 0
 #define ANET_IP_ONLY (1<<0)
 
-#if defined(__sun)
+#if defined(__sun) || defined(_AIX)
 #define AF_LOCAL AF_UNIX
 #endif
 
+#ifdef _AIX
+#undef ip_len
+#endif
+
 int anetTcpConnect(char *err, char *addr, int port);
 int anetTcpNonBlockConnect(char *err, char *addr, int port);
 int anetTcpNonBlockBindConnect(char *err, char *addr, int port, char *source_addr);
@@ -58,9 +62,11 @@ int anetTcpAccept(char *err, int serversock, char *ip, size_t ip_len, int *port)
 int anetUnixAccept(char *err, int serversock);
 int anetWrite(int fd, char *buf, int count);
 int anetNonBlock(char *err, int fd);
+int anetBlock(char *err, int fd);
 int anetEnableTcpNoDelay(char *err, int fd);
 int anetDisableTcpNoDelay(char *err, int fd);
 int anetTcpKeepAlive(char *err, int fd);
+int anetSendTimeout(char *err, int fd, long long ms);
 int anetPeerToString(int fd, char *ip, size_t ip_len, int *port);
 int anetKeepAlive(char *err, int fd, int interval);
 int anetSockName(int fd, char *ip, size_t ip_len, int *port);

src/aof.c

@@ -74,7 +74,7 @@ void aofRewriteBufferReset(void) {
     listSetFreeMethod(server.aof_rewrite_buf_blocks,zfree);
 }
 
-/* Return the current size of the AOF rerwite buffer. */
+/* Return the current size of the AOF rewrite buffer. */
 unsigned long aofRewriteBufferSize(void) {
     listNode *ln;
     listIter li;
@@ -95,6 +95,10 @@ void aofChildWriteDiffData(aeEventLoop *el, int fd, void *privdata, int mask) {
     listNode *ln;
     aofrwblock *block;
     ssize_t nwritten;
+    REDIS_NOTUSED(el);
+    REDIS_NOTUSED(fd);
+    REDIS_NOTUSED(privdata);
+    REDIS_NOTUSED(mask);
 
     while(1) {
         ln = listFirst(server.aof_rewrite_buf_blocks);
@@ -177,7 +181,7 @@ ssize_t aofRewriteBufferWrite(int fd) {
 
         if (block->used) {
             nwritten = write(fd,block->buf,block->used);
-            if (nwritten != block->used) {
+            if (nwritten != (ssize_t)block->used) {
                 if (nwritten == 0) errno = EIO;
                 return -1;
             }
@@ -241,7 +245,7 @@ int startAppendOnly(void) {
         redisLog(REDIS_WARNING,"Redis needs to enable the AOF but can't trigger a background AOF rewrite operation. Check the above logs for more info about the error.");
         return REDIS_ERR;
     }
-    /* We correctly switched on AOF, now wait for the rerwite to be complete
+    /* We correctly switched on AOF, now wait for the rewrite to be complete
      * in order to append data on disk. */
     server.aof_state = REDIS_AOF_WAIT_REWRITE;
     return REDIS_OK;
@@ -282,7 +286,7 @@ void flushAppendOnlyFile(int force) {
          * the write for a couple of seconds. */
         if (sync_in_progress) {
             if (server.aof_flush_postponed_start == 0) {
-                /* No previous write postponinig, remember that we are
+                /* No previous write postponing, remember that we are
                  * postponing the flush and return. */
                 server.aof_flush_postponed_start = server.unixtime;
                 return;
@@ -333,7 +337,7 @@ void flushAppendOnlyFile(int force) {
             last_write_error_log = server.unixtime;
         }
 
-        /* Lof the AOF write error and record the error code. */
+        /* Log the AOF write error and record the error code. */
         if (nwritten == -1) {
             if (can_log) {
                 redisLog(REDIS_WARNING,"Error writing to the AOF file: %s",
@@ -357,7 +361,7 @@ void flushAppendOnlyFile(int force) {
                              "ftruncate: %s", strerror(errno));
                 }
             } else {
-                /* If the ftrunacate() succeeded we can set nwritten to
+                /* If the ftruncate() succeeded we can set nwritten to
                  * -1 since there is no longer partial data into the AOF. */
                 nwritten = -1;
             }
@@ -369,7 +373,7 @@ void flushAppendOnlyFile(int force) {
             /* We can't recover when the fsync policy is ALWAYS since the
              * reply for the client is already in the output buffers, and we
              * have the contract with the user that on acknowledged write data
-             * is synched on disk. */
+             * is synced on disk. */
             redisLog(REDIS_WARNING,"Can't recover from AOF write error when the AOF fsync policy is 'always'. Exiting...");
             exit(1);
         } else {
@@ -464,7 +468,7 @@ sds catAppendOnlyExpireAtCommand(sds buf, struct redisCommand *cmd, robj *key, r
     long long when;
     robj *argv[3];
 
-    /* Make sure we can use strtol */
+    /* Make sure we can use strtoll */
     seconds = getDecodedObject(seconds);
     when = strtoll(seconds->ptr,NULL,10);
     /* Convert argument into milliseconds for EXPIRE, SETEX, EXPIREAT */
@@ -495,7 +499,7 @@ void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv, int a
     robj *tmpargv[3];
 
     /* The DB this command was targeting is not the same as the last command
-     * we appendend. To issue a SELECT command is needed. */
+     * we appended. To issue a SELECT command is needed. */
     if (dictid != server.aof_selected_db) {
         char seldb[64];
 
@@ -573,6 +577,14 @@ struct redisClient *createFakeClient(void) {
     return c;
 }
 
+void freeFakeClientArgv(struct redisClient *c) {
+    int j;
+
+    for (j = 0; j < c->argc; j++)
+        decrRefCount(c->argv[j]);
+    zfree(c->argv);
+}
+
 void freeFakeClient(struct redisClient *c) {
     sdsfree(c->querybuf);
     listRelease(c->reply);
@@ -581,7 +593,7 @@ void freeFakeClient(struct redisClient *c) {
     zfree(c);
 }
 
-/* Replay the append log file. On error REDIS_OK is returned. On non fatal
+/* Replay the append log file. On success REDIS_OK is returned. On non fatal
  * error (the append only file is zero-length) REDIS_ERR is returned. On
  * fatal error an error message is logged and the program exists. */
 int loadAppendOnlyFile(char *filename) {
@@ -590,6 +602,7 @@ int loadAppendOnlyFile(char *filename) {
     struct redis_stat sb;
     int old_aof_state = server.aof_state;
     long loops = 0;
+    off_t valid_up_to = 0; /* Offset of the latest well-formed command loaded. */
 
     if (fp && redis_fstat(fileno(fp),&sb) != -1 && sb.st_size == 0) {
         server.aof_current_size = 0;
@@ -630,18 +643,35 @@ int loadAppendOnlyFile(char *filename) {
                 goto readerr;
         }
         if (buf[0] != '*') goto fmterr;
+        if (buf[1] == '\0') goto readerr;
         argc = atoi(buf+1);
         if (argc < 1) goto fmterr;
 
         argv = zmalloc(sizeof(robj*)*argc);
+        fakeClient->argc = argc;
+        fakeClient->argv = argv;
+
         for (j = 0; j < argc; j++) {
-            if (fgets(buf,sizeof(buf),fp) == NULL) goto readerr;
+            if (fgets(buf,sizeof(buf),fp) == NULL) {
+                fakeClient->argc = j; /* Free up to j-1. */
+                freeFakeClientArgv(fakeClient);
+                goto readerr;
+            }
             if (buf[0] != '$') goto fmterr;
             len = strtol(buf+1,NULL,10);
             argsds = sdsnewlen(NULL,len);
-            if (len && fread(argsds,len,1,fp) == 0) goto fmterr;
+            if (len && fread(argsds,len,1,fp) == 0) {
+                sdsfree(argsds);
+                fakeClient->argc = j; /* Free up to j-1. */
+                freeFakeClientArgv(fakeClient);
+                goto readerr;
+            }
             argv[j] = createObject(REDIS_STRING,argsds);
-            if (fread(buf,2,1,fp) == 0) goto fmterr; /* discard CRLF */
+            if (fread(buf,2,1,fp) == 0) {
+                fakeClient->argc = j+1; /* Free up to j. */
+                freeFakeClientArgv(fakeClient);
+                goto readerr; /* discard CRLF */
+            }
         }
 
         /* Command lookup */
@@ -650,9 +680,8 @@ int loadAppendOnlyFile(char *filename) {
             redisLog(REDIS_WARNING,"Unknown command '%s' reading the append only file", (char*)argv[0]->ptr);
             exit(1);
         }
+
         /* Run the command in the context of a fake client */
-        fakeClient->argc = argc;
-        fakeClient->argv = argv;
         cmd->proc(fakeClient);
 
         /* The fake client should not have a reply */
@@ -662,15 +691,15 @@ int loadAppendOnlyFile(char *filename) {
 
         /* Clean up. Command code may have changed argv/argc so we use the
          * argv/argc of the client instead of the local variables. */
-        for (j = 0; j < fakeClient->argc; j++)
-            decrRefCount(fakeClient->argv[j]);
-        zfree(fakeClient->argv);
+        freeFakeClientArgv(fakeClient);
+        if (server.aof_load_truncated) valid_up_to = ftello(fp);
     }
 
     /* This point can only be reached when EOF is reached without errors.
      * If the client is in the middle of a MULTI/EXEC, log error and quit. */
-    if (fakeClient->flags & REDIS_MULTI) goto readerr;
+    if (fakeClient->flags & REDIS_MULTI) goto uxeof;
 
+loaded_ok: /* DB loaded, cleanup and return REDIS_OK to the caller. */
     fclose(fp);
     freeFakeClient(fakeClient);
     server.aof_state = old_aof_state;
@@ -679,14 +708,41 @@ int loadAppendOnlyFile(char *filename) {
     server.aof_rewrite_base_size = server.aof_current_size;
     return REDIS_OK;
 
-readerr:
-    if (feof(fp)) {
-        redisLog(REDIS_WARNING,"Unexpected end of file reading the append only file");
-    } else {
+readerr: /* Read error. If feof(fp) is true, fall through to unexpected EOF. */
+    if (!feof(fp)) {
         redisLog(REDIS_WARNING,"Unrecoverable error reading the append only file: %s", strerror(errno));
+        exit(1);
     }
+
+uxeof: /* Unexpected AOF end of file. */
+    if (server.aof_load_truncated) {
+        redisLog(REDIS_WARNING,"!!! Warning: short read while loading the AOF file !!!");
+        redisLog(REDIS_WARNING,"!!! Truncating the AOF at offset %llu !!!",
+            (unsigned long long) valid_up_to);
+        if (valid_up_to == -1 || truncate(filename,valid_up_to) == -1) {
+            if (valid_up_to == -1) {
+                redisLog(REDIS_WARNING,"Last valid command offset is invalid");
+            } else {
+                redisLog(REDIS_WARNING,"Error truncating the AOF file: %s",
+                    strerror(errno));
+            }
+        } else {
+            /* Make sure the AOF file descriptor points to the end of the
+             * file after the truncate call. */
+            if (server.aof_fd != -1 && lseek(server.aof_fd,0,SEEK_END) == -1) {
+                redisLog(REDIS_WARNING,"Can't seek the end of the AOF file: %s",
+                    strerror(errno));
+            } else {
+                redisLog(REDIS_WARNING,
+                    "AOF loaded anyway because aof-load-truncated is enabled");
+                goto loaded_ok;
+            }
+        }
+    }
+    redisLog(REDIS_WARNING,"Unexpected end of file reading the append only file. You can: 1) Make a backup of your AOF file, then use ./redis-check-aof --fix <filename>. 2) Alternatively you can set the 'aof-load-truncated' configuration option to yes and restart the server.");
     exit(1);
-fmterr:
+
+fmterr: /* Format error. */
     redisLog(REDIS_WARNING,"Bad file format reading the append only file: make a backup of your AOF file, then use ./redis-check-aof --fix <filename>");
     exit(1);
 }
@@ -944,7 +1000,7 @@ int rewriteHashObject(rio *r, robj *key, robj *o) {
  * the difference accumulated from the parent into a buffer, that is
  * concatenated at the end of the rewrite. */
 ssize_t aofReadDiffFromParent(void) {
-    char buf[65536]; /* Default pipe buffer size on most Linux sytems. */
+    char buf[65536]; /* Default pipe buffer size on most Linux systems. */
     ssize_t nread, total = 0;
 
     while ((nread =
@@ -1058,7 +1114,7 @@ int rewriteAppendOnlyFile(char *filename) {
 
     /* Read again a few times to get more data from the parent.
      * We can't read forever (the server may receive data from clients
-     * fater than it is able to send data to the child), so we try to read
+     * faster than it is able to send data to the child), so we try to read
      * some more data in a loop as soon as there is a good chance more data
      * will come. If it looks like we are wasting time, we abort (this
      * happens after 20 ms without new data). */
@@ -1128,6 +1184,9 @@ werr:
  * parent sends a '!' as well to acknowledge. */
 void aofChildPipeReadable(aeEventLoop *el, int fd, void *privdata, int mask) {
     char byte;
+    REDIS_NOTUSED(el);
+    REDIS_NOTUSED(privdata);
+    REDIS_NOTUSED(mask);
 
     if (read(fd,&byte,1) == 1 && byte == '!') {
         redisLog(REDIS_NOTICE,"AOF rewrite child asks to stop sending diffs.");
@@ -1191,7 +1250,7 @@ void aofClosePipes(void) {
 }
 
 /* ----------------------------------------------------------------------------
- * AOF backgorund rewrite
+ * AOF background rewrite
  * ------------------------------------------------------------------------- */
 
 /* This is how rewriting of the append only file in background works:
@@ -1280,7 +1339,7 @@ void aofRemoveTempFile(pid_t childpid) {
     unlink(tmpfile);
 }
 
-/* Update the server.aof_current_size filed explicitly using stat(2)
+/* Update the server.aof_current_size field explicitly using stat(2)
  * to check the size of the file. This is useful after a rewrite or after
  * a restart, normally the size is updated just adding the write length
  * to the current length, that is much faster. */
@@ -1333,7 +1392,7 @@ void backgroundRewriteDoneHandler(int exitcode, int bysignal) {
         latencyAddSampleIfNeeded("aof-rewrite-diff-write",latency);
 
         redisLog(REDIS_NOTICE,
-            "Redidual parent diff successfully flushed to the rewritten AOF (%.2f MB)", (double) aofRewriteBufferSize() / (1024*1024));
+            "Residual parent diff successfully flushed to the rewritten AOF (%.2f MB)", (double) aofRewriteBufferSize() / (1024*1024));
 
         /* The only remaining thing to do is to rename the temporary file to
          * the configured file and switch the file descriptor used to do AOF

src/bitops.c

@@ -107,12 +107,12 @@ size_t redisPopcount(void *s, long count) {
  * no zero bit is found, it returns count*8 assuming the string is zero
  * padded on the right. However if 'bit' is 1 it is possible that there is
  * not a single set bit in the bitmap. In this special case -1 is returned. */
-long redisBitpos(void *s, long count, int bit) {
+long redisBitpos(void *s, unsigned long count, int bit) {
     unsigned long *l;
     unsigned char *c;
     unsigned long skipval, word = 0, one;
     long pos = 0; /* Position of bit, to return to the caller. */
-    int j;
+    unsigned long j;
 
     /* Process whole words first, seeking for first word that is not
      * all ones or all zeros respectively if we are lookig for zeros
@@ -276,11 +276,12 @@ void getbitCommand(redisClient *c) {
 void bitopCommand(redisClient *c) {
     char *opname = c->argv[1]->ptr;
     robj *o, *targetkey = c->argv[2];
-    long op, j, numkeys;
+    unsigned long op, j, numkeys;
     robj **objects;      /* Array of source objects. */
     unsigned char **src; /* Array of source strings pointers. */
-    long *len, maxlen = 0; /* Array of length of src strings, and max len. */
-    long minlen = 0;    /* Min len among the input keys. */
+    unsigned long *len, maxlen = 0; /* Array of length of src strings,
+                                       and max len. */
+    unsigned long minlen = 0;    /* Min len among the input keys. */
     unsigned char *res = NULL; /* Resulting string. */
 
     /* Parse the operation name. */
@@ -320,9 +321,10 @@ void bitopCommand(redisClient *c) {
         }
         /* Return an error if one of the keys is not a string. */
         if (checkType(c,o,REDIS_STRING)) {
-            for (j = j-1; j >= 0; j--) {
-                if (objects[j])
-                    decrRefCount(objects[j]);
+            unsigned long i;
+            for (i = 0; i < j; i++) {
+                if (objects[i])
+                    decrRefCount(objects[i]);
             }
             zfree(src);
             zfree(len);
@@ -340,7 +342,7 @@ void bitopCommand(redisClient *c) {
     if (maxlen) {
         res = (unsigned char*) sdsnewlen(NULL,maxlen);
         unsigned char output, byte;
-        long i;
+        unsigned long i;
 
         /* Fast path: as far as we have data for all the input bitmaps we
          * can take a fast path that performs much better than the

src/cluster.c

@@ -72,6 +72,7 @@ void resetManualFailover(void);
 void clusterCloseAllSlots(void);
 void clusterSetNodeAsMaster(clusterNode *n);
 void clusterDelNode(clusterNode *delnode);
+sds representRedisNodeFlags(sds ci, uint16_t flags);
 
 /* -----------------------------------------------------------------------------
  * Initialization
@@ -123,7 +124,7 @@ int clusterLoadConfig(char *filename) {
         return REDIS_ERR;
     }
 
-    /* Parse the file. Note that single liens of the cluster config file can
+    /* Parse the file. Note that single lines of the cluster config file can
      * be really long as they include all the hash slots of the node.
      * This means in the worst possible case, half of the Redis slots will be
      * present in a single line, possibly in importing or migrating state, so
@@ -163,9 +164,13 @@ int clusterLoadConfig(char *filename) {
                         argv[j]);
                 }
             }
+            sdsfreesplitres(argv,argc);
             continue;
         }
 
+        /* Regular config lines have at least eight fields */
+        if (argc < 8) goto fmterr;
+
         /* Create this node if it does not exist */
         n = clusterLookupNode(argv[0]);
         if (!n) {
@@ -173,7 +178,7 @@ int clusterLoadConfig(char *filename) {
             clusterAddNode(n);
         }
         /* Address and port */
-        if ((p = strchr(argv[1],':')) == NULL) goto fmterr;
+        if ((p = strrchr(argv[1],':')) == NULL) goto fmterr;
         *p = '\0';
         memcpy(n->ip,argv[1],strlen(argv[1])+1);
         n->port = atoi(p+1);
@@ -266,11 +271,12 @@ int clusterLoadConfig(char *filename) {
 
         sdsfreesplitres(argv,argc);
     }
+    /* Config sanity check */
+    if (server.cluster->myself == NULL) goto fmterr;
+
     zfree(line);
     fclose(fp);
 
-    /* Config sanity check */
-    redisAssert(server.cluster->myself != NULL);
     redisLog(REDIS_NOTICE,"Node configuration loaded, I'm %.40s", myself->name);
 
     /* Something that should never happen: currentEpoch smaller than
@@ -284,7 +290,8 @@ int clusterLoadConfig(char *filename) {
 fmterr:
     redisLog(REDIS_WARNING,
         "Unrecoverable error: corrupted cluster config file.");
-    fclose(fp);
+    zfree(line);
+    if (fp) fclose(fp);
     exit(1);
 }
 
@@ -321,7 +328,7 @@ int clusterSaveConfig(int do_fsync) {
 
     /* Pad the new payload if the existing file length is greater. */
     if (fstat(fd,&sb) != -1) {
-        if (sb.st_size > content_size) {
+        if (sb.st_size > (off_t)content_size) {
             ci = sdsgrowzero(ci,sb.st_size);
             memset(ci+content_size,'\n',sb.st_size-content_size);
         }
@@ -410,6 +417,7 @@ void clusterInit(void) {
     server.cluster->failover_auth_count = 0;
     server.cluster->failover_auth_rank = 0;
     server.cluster->failover_auth_epoch = 0;
+    server.cluster->cant_failover_reason = REDIS_CLUSTER_CANT_FAILOVER_NONE;
     server.cluster->lastVoteEpoch = 0;
     server.cluster->stats_bus_messages_sent = 0;
     server.cluster->stats_bus_messages_received = 0;
@@ -481,7 +489,8 @@ void clusterInit(void) {
  * 3) If the node is a slave, it turns into a master.
  * 5) Only for hard reset: a new Node ID is generated.
  * 6) Only for hard reset: currentEpoch and configEpoch are set to 0.
- * 7) The new configuration is saved and the cluster state updated.  */
+ * 7) The new configuration is saved and the cluster state updated.
+ * 8) If the node was a slave, the whole data set is flushed away. */
 void clusterReset(int hard) {
     dictIterator *di;
     dictEntry *de;
@@ -491,6 +500,7 @@ void clusterReset(int hard) {
     if (nodeIsSlave(myself)) {
         clusterSetNodeAsMaster(myself);
         replicationUnsetMaster();
+        emptyDb(NULL);
     }
 
     /* Close slots, reset manual failover state. */
@@ -1124,7 +1134,7 @@ int clusterStartHandshake(char *ip, int port) {
 
     /* Add the node with a random address (NULL as first argument to
      * createClusterNode()). Everything will be fixed during the
-     * handskake. */
+     * handshake. */
     n = createClusterNode(NULL,REDIS_NODE_HANDSHAKE|REDIS_NODE_MEET);
     memcpy(n->ip,norm_ip,sizeof(n->ip));
     n->port = port;
@@ -1142,20 +1152,11 @@ void clusterProcessGossipSection(clusterMsg *hdr, clusterLink *link) {
     clusterNode *sender = link->node ? link->node : clusterLookupNode(hdr->sender);
 
     while(count--) {
-        sds ci = sdsempty();
         uint16_t flags = ntohs(g->flags);
         clusterNode *node;
+        sds ci;
 
-        if (flags == 0) ci = sdscat(ci,"noflags,");
-        if (flags & REDIS_NODE_MYSELF) ci = sdscat(ci,"myself,");
-        if (flags & REDIS_NODE_MASTER) ci = sdscat(ci,"master,");
-        if (flags & REDIS_NODE_SLAVE) ci = sdscat(ci,"slave,");
-        if (flags & REDIS_NODE_PFAIL) ci = sdscat(ci,"fail?,");
-        if (flags & REDIS_NODE_FAIL) ci = sdscat(ci,"fail,");
-        if (flags & REDIS_NODE_HANDSHAKE) ci = sdscat(ci,"handshake,");
-        if (flags & REDIS_NODE_NOADDR) ci = sdscat(ci,"noaddr,");
-        if (ci[sdslen(ci)-1] == ',') ci[sdslen(ci)-1] = ' ';
-
+        ci = representRedisNodeFlags(sdsempty(), flags);
         redisLog(REDIS_DEBUG,"GOSSIP %.40s %s:%d %s",
             g->nodename,
             g->ip,
@@ -1284,7 +1285,7 @@ void clusterSetNodeAsMaster(clusterNode *n) {
  * node (see the function comments for more info).
  *
  * The 'sender' is the node for which we received a configuration update.
- * Sometimes it is not actaully the "Sender" of the information, like in the case
+ * Sometimes it is not actually the "Sender" of the information, like in the case
  * we receive the info via an UPDATE packet. */
 void clusterUpdateSlotsConfigWith(clusterNode *sender, uint64_t senderConfigEpoch, unsigned char *slots) {
     int j;
@@ -1481,7 +1482,8 @@ int clusterProcessPacket(clusterLink *link) {
     } else if (type == CLUSTERMSG_TYPE_PUBLISH) {
         uint32_t explen = sizeof(clusterMsg)-sizeof(union clusterMsgData);
 
-        explen += sizeof(clusterMsgDataPublish) +
+        explen += sizeof(clusterMsgDataPublish) -
+                8 +
                 ntohl(hdr->data.publish.msg.channel_len) +
                 ntohl(hdr->data.publish.msg.message_len);
         if (totlen != explen) return 1;
@@ -1532,7 +1534,7 @@ int clusterProcessPacket(clusterLink *link) {
         }
     }
 
-    /* Process packets by type. */
+    /* Initial processing of PING and MEET requests replying with a PONG. */
     if (type == CLUSTERMSG_TYPE_PING || type == CLUSTERMSG_TYPE_MEET) {
         redisLog(REDIS_DEBUG,"Ping packet received: %p", (void*)link->node);
 
@@ -1570,14 +1572,17 @@ int clusterProcessPacket(clusterLink *link) {
             clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG);
         }
 
-        /* Get info from the gossip section */
-        clusterProcessGossipSection(hdr,link);
+        /* If this is a MEET packet from an unknown node, we still process
+         * the gossip section here since we have to trust the sender because
+         * of the message type. */
+        if (!sender && type == CLUSTERMSG_TYPE_MEET)
+            clusterProcessGossipSection(hdr,link);
 
         /* Anyway reply with a PONG */
         clusterSendPing(link,CLUSTERMSG_TYPE_PONG);
     }
 
-    /* PING or PONG: process config information. */
+    /* PING, PONG, MEET: process config information. */
     if (type == CLUSTERMSG_TYPE_PING || type == CLUSTERMSG_TYPE_PONG ||
         type == CLUSTERMSG_TYPE_MEET)
     {
@@ -1597,7 +1602,7 @@ int clusterProcessPacket(clusterLink *link) {
                         clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG|
                                              CLUSTER_TODO_UPDATE_STATE);
                     }
-                    /* Free this node as we alrady have it. This will
+                    /* Free this node as we already have it. This will
                      * cause the link to be freed as well. */
                     freeClusterNode(link->node);
                     return 0;
@@ -1774,7 +1779,7 @@ int clusterProcessPacket(clusterLink *link) {
         }
 
         /* Get info from the gossip section */
-        clusterProcessGossipSection(hdr,link);
+        if (sender) clusterProcessGossipSection(hdr,link);
     } else if (type == CLUSTERMSG_TYPE_FAIL) {
         clusterNode *failing;
 
@@ -1794,7 +1799,7 @@ int clusterProcessPacket(clusterLink *link) {
             }
         } else {
             redisLog(REDIS_NOTICE,
-                "Ignoring FAIL message from unknonw node %.40s about %.40s",
+                "Ignoring FAIL message from unknown node %.40s about %.40s",
                 hdr->sender, hdr->data.fail.about.nodename);
         }
     } else if (type == CLUSTERMSG_TYPE_PUBLISH) {
@@ -1863,7 +1868,7 @@ int clusterProcessPacket(clusterLink *link) {
         clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG|
                              CLUSTER_TODO_FSYNC_CONFIG);
 
-        /* Check the bitmap of served slots and udpate our
+        /* Check the bitmap of served slots and update our
          * config accordingly. */
         clusterUpdateSlotsConfigWith(n,reportedConfigEpoch,
             hdr->data.update.nodecfg.slots);
@@ -1912,7 +1917,7 @@ void clusterReadHandler(aeEventLoop *el, int fd, void *privdata, int mask) {
     ssize_t nread;
     clusterMsg *hdr;
     clusterLink *link = (clusterLink*) privdata;
-    int readlen, rcvbuflen;
+    unsigned int readlen, rcvbuflen;
     REDIS_NOTUSED(el);
     REDIS_NOTUSED(mask);
 
@@ -2180,7 +2185,7 @@ void clusterSendPublish(clusterLink *link, robj *channel, robj *message) {
 
     clusterBuildMessageHdr(hdr,CLUSTERMSG_TYPE_PUBLISH);
     totlen = sizeof(clusterMsg)-sizeof(union clusterMsgData);
-    totlen += sizeof(clusterMsgDataPublish) + channel_len + message_len;
+    totlen += sizeof(clusterMsgDataPublish) - 8 + channel_len + message_len;
 
     hdr->data.publish.msg.channel_len = htonl(channel_len);
     hdr->data.publish.msg.message_len = htonl(message_len);
@@ -2430,6 +2435,68 @@ int clusterGetSlaveRank(void) {
     return rank;
 }
 
+/* This function is called by clusterHandleSlaveFailover() in order to
+ * let the slave log why it is not able to failover. Sometimes there are
+ * not the conditions, but since the failover function is called again and
+ * again, we can't log the same things continuously.
+ *
+ * This function works by logging only if a given set of conditions are
+ * true:
+ *
+ * 1) The reason for which the failover can't be initiated changed.
+ *    The reasons also include a NONE reason we reset the state to
+ *    when the slave finds that its master is fine (no FAIL flag).
+ * 2) Also, the log is emitted again if the master is still down and
+ *    the reason for not failing over is still the same, but more than
+ *    REDIS_CLUSTER_CANT_FAILOVER_RELOG_PERIOD seconds elapsed.
+ * 3) Finally, the function only logs if the slave is down for more than
+ *    five seconds + NODE_TIMEOUT. This way nothing is logged when a
+ *    failover starts in a reasonable time.
+ *
+ * The function is called with the reason why the slave can't failover
+ * which is one of the integer macros REDIS_CLUSTER_CANT_FAILOVER_*.
+ *
+ * The function is guaranteed to be called only if 'myself' is a slave. */
+void clusterLogCantFailover(int reason) {
+    char *msg;
+    static time_t lastlog_time = 0;
+    mstime_t nolog_fail_time = server.cluster_node_timeout + 5000;
+
+    /* Don't log if we have the same reason for some time. */
+    if (reason == server.cluster->cant_failover_reason &&
+        time(NULL)-lastlog_time < REDIS_CLUSTER_CANT_FAILOVER_RELOG_PERIOD)
+        return;
+
+    server.cluster->cant_failover_reason = reason;
+
+    /* We also don't emit any log if the master failed no long ago, the
+     * goal of this function is to log slaves in a stalled condition for
+     * a long time. */
+    if (myself->slaveof &&
+        nodeFailed(myself->slaveof) &&
+        (mstime() - myself->slaveof->fail_time) < nolog_fail_time) return;
+
+    switch(reason) {
+    case REDIS_CLUSTER_CANT_FAILOVER_DATA_AGE:
+        msg = "Disconnected from master for longer than allowed.";
+        break;
+    case REDIS_CLUSTER_CANT_FAILOVER_WAITING_DELAY:
+        msg = "Waiting the delay before I can start a new failover.";
+        break;
+    case REDIS_CLUSTER_CANT_FAILOVER_EXPIRED:
+        msg = "Failover attempt expired.";
+        break;
+    case REDIS_CLUSTER_CANT_FAILOVER_WAITING_VOTES:
+        msg = "Waiting for votes, but majority still not reached.";
+        break;
+    default:
+        msg = "Unknown reason code.";
+        break;
+    }
+    lastlog_time = time(NULL);
+    redisLog(REDIS_WARNING,"Currently unable to failover: %s", msg);
+}
+
 /* This function is called if we are a slave node and our master serving
  * a non-zero amount of hash slots is in FAIL state.
  *
@@ -2468,7 +2535,13 @@ void clusterHandleSlaveFailover(void) {
     if (nodeIsMaster(myself) ||
         myself->slaveof == NULL ||
         (!nodeFailed(myself->slaveof) && !manual_failover) ||
-        myself->slaveof->numslots == 0) return;
+        myself->slaveof->numslots == 0)
+    {
+        /* There are no reasons to failover, so we set the reason why we
+         * are returning without failing over to NONE. */
+        server.cluster->cant_failover_reason = REDIS_CLUSTER_CANT_FAILOVER_NONE;
+        return;
+    }
 
     /* Set data_age to the number of seconds we are disconnected from
      * the master. */
@@ -2494,7 +2567,10 @@ void clusterHandleSlaveFailover(void) {
         (((mstime_t)server.repl_ping_slave_period * 1000) +
          (server.cluster_node_timeout * server.cluster_slave_validity_factor)))
     {
-        if (!manual_failover) return;
+        if (!manual_failover) {
+            clusterLogCantFailover(REDIS_CLUSTER_CANT_FAILOVER_DATA_AGE);
+            return;
+        }
     }
 
     /* If the previous failover attempt timedout and the retry time has
@@ -2550,10 +2626,16 @@ void clusterHandleSlaveFailover(void) {
     }
 
     /* Return ASAP if we can't still start the election. */
-    if (mstime() < server.cluster->failover_auth_time) return;
+    if (mstime() < server.cluster->failover_auth_time) {
+        clusterLogCantFailover(REDIS_CLUSTER_CANT_FAILOVER_WAITING_DELAY);
+        return;
+    }
 
     /* Return ASAP if the election is too old to be valid. */
-    if (auth_age > auth_timeout) return;
+    if (auth_age > auth_timeout) {
+        clusterLogCantFailover(REDIS_CLUSTER_CANT_FAILOVER_EXPIRED);
+        return;
+    }
 
     /* Ask for votes if needed. */
     if (server.cluster->failover_auth_sent == 0) {
@@ -2608,6 +2690,8 @@ void clusterHandleSlaveFailover(void) {
 
         /* 6) If there was a manual failover in progress, clear the state. */
         resetManualFailover();
+    } else {
+        clusterLogCantFailover(REDIS_CLUSTER_CANT_FAILOVER_WAITING_VOTES);
     }
 }
 
@@ -2672,6 +2756,10 @@ void clusterHandleSlaveMigration(int max_slaves) {
 
         /* Only iterate over working masters. */
         if (nodeIsSlave(node) || nodeFailed(node)) continue;
+        /* If this master never had slaves so far, don't migrate. We want
+         * to migrate to a master that remained orphaned, not masters that
+         * were never configured to have slaves. */
+        if (node->numslaves == 0) continue;
         okslaves = clusterCountNonFailingSlaves(node);
 
         if (okslaves == 0 && target == NULL && node->numslots > 0)
@@ -2757,7 +2845,7 @@ void clusterHandleManualFailover(void) {
     /* Return ASAP if no manual failover is in progress. */
     if (server.cluster->mf_end == 0) return;
 
-    /* If mf_can_start is non-zero, the failover was alrady triggered so the
+    /* If mf_can_start is non-zero, the failover was already triggered so the
      * next steps are performed by clusterHandleSlaveFailover(). */
     if (server.cluster->mf_can_start) return;
 
@@ -2821,6 +2909,12 @@ void clusterCron(void) {
             fd = anetTcpNonBlockBindConnect(server.neterr, node->ip,
                 node->port+REDIS_CLUSTER_PORT_INCR, REDIS_BIND_ADDR);
             if (fd == -1) {
+                /* We got a synchronous error from connect before
+                 * clusterSendPing() had a chance to be called.
+                 * If node->ping_sent is zero, failure detection can't work,
+                 * so we claim we actually sent a ping now (that will
+                 * be really sent as soon as the link is obtained). */
+                if (node->ping_sent == 0) node->ping_sent = mstime();
                 redisLog(REDIS_DEBUG, "Unable to connect to "
                     "Cluster Node [%s]:%d -> %s", node->ip,
                     node->port+REDIS_CLUSTER_PORT_INCR,
@@ -2910,7 +3004,11 @@ void clusterCron(void) {
         if (nodeIsSlave(myself) && nodeIsMaster(node) && !nodeFailed(node)) {
             int okslaves = clusterCountNonFailingSlaves(node);
 
-            if (okslaves == 0 && node->numslots > 0) orphaned_masters++;
+            /* A master is orphaned if it is serving a non-zero number of
+             * slots, have no working slaves, but used to have at least one
+             * slave. */
+            if (okslaves == 0 && node->numslots > 0 && node->numslaves)
+                orphaned_masters++;
             if (okslaves > max_slaves) max_slaves = okslaves;
             if (nodeIsSlave(myself) && myself->slaveof == node)
                 this_slaves = okslaves;
@@ -3141,7 +3239,7 @@ void clusterCloseAllSlots(void) {
 
 void clusterUpdateState(void) {
     int j, new_state;
-    int unreachable_masters = 0;
+    int reachable_masters = 0;
     static mstime_t among_minority_time;
     static mstime_t first_call_time = 0;
 
@@ -3163,20 +3261,22 @@ void clusterUpdateState(void) {
     new_state = REDIS_CLUSTER_OK;
 
     /* Check if all the slots are covered. */
-    for (j = 0; j < REDIS_CLUSTER_SLOTS; j++) {
-        if (server.cluster->slots[j] == NULL ||
-            server.cluster->slots[j]->flags & (REDIS_NODE_FAIL))
-        {
-            new_state = REDIS_CLUSTER_FAIL;
-            break;
+    if (server.cluster_require_full_coverage) {
+        for (j = 0; j < REDIS_CLUSTER_SLOTS; j++) {
+            if (server.cluster->slots[j] == NULL ||
+                server.cluster->slots[j]->flags & (REDIS_NODE_FAIL))
+            {
+                new_state = REDIS_CLUSTER_FAIL;
+                break;
+            }
         }
     }
 
     /* Compute the cluster size, that is the number of master nodes
      * serving at least a single slot.
      *
-     * At the same time count the number of unreachable masters with
-     * at least one node. */
+     * At the same time count the number of reachable masters having
+     * at least one slot. */
     {
         dictIterator *di;
         dictEntry *de;
@@ -3188,20 +3288,19 @@ void clusterUpdateState(void) {
 
             if (nodeIsMaster(node) && node->numslots) {
                 server.cluster->size++;
-                if (node->flags & (REDIS_NODE_FAIL|REDIS_NODE_PFAIL))
-                    unreachable_masters++;
+                if ((node->flags & (REDIS_NODE_FAIL|REDIS_NODE_PFAIL)) == 0)
+                    reachable_masters++;
             }
         }
         dictReleaseIterator(di);
     }
 
-    /* If we can't reach at least half the masters, change the cluster state
-     * to FAIL, as we are not even able to mark nodes as FAIL in this side
-     * of the netsplit because of lack of majority. */
+    /* If we are in a minority partition, change the cluster state
+     * to FAIL. */
     {
         int needed_quorum = (server.cluster->size / 2) + 1;
 
-        if (unreachable_masters >= needed_quorum) {
+        if (reachable_masters < needed_quorum) {
             new_state = REDIS_CLUSTER_FAIL;
             among_minority_time = mstime();
         }
@@ -3284,16 +3383,15 @@ int verifyClusterConfigWithData(void) {
          * assigned to this slot. Fix this condition. */
 
         update_config++;
-        /* Case A: slot is unassigned. Take responsability for it. */
+        /* Case A: slot is unassigned. Take responsibility for it. */
         if (server.cluster->slots[j] == NULL) {
-            redisLog(REDIS_WARNING, "I've keys about slot %d that is "
-                                    "unassigned. Taking responsability "
-                                    "for it.",j);
+            redisLog(REDIS_WARNING, "I have keys for unassigned slot %d. "
+                                    "Taking responsibility for it.",j);
             clusterAddSlot(myself,j);
         } else {
-            redisLog(REDIS_WARNING, "I've keys about slot %d that is "
-                                    "already assigned to a different node. "
-                                    "Setting it in importing state.",j);
+            redisLog(REDIS_WARNING, "I have keys for slot %d, but the slot is "
+                                    "assigned to another node. "
+                                    "Setting it to importing state.",j);
             server.cluster->importing_slots_from[j] = server.cluster->slots[j];
         }
     }
@@ -3326,9 +3424,40 @@ void clusterSetMaster(clusterNode *n) {
 }
 
 /* -----------------------------------------------------------------------------
- * CLUSTER command
+ * Nodes to string representation functions.
  * -------------------------------------------------------------------------- */
 
+struct redisNodeFlags {
+    uint16_t flag;
+    char *name;
+};
+
+static struct redisNodeFlags redisNodeFlagsTable[] = {
+    {REDIS_NODE_MYSELF,    "myself,"},
+    {REDIS_NODE_MASTER,    "master,"},
+    {REDIS_NODE_SLAVE,     "slave,"},
+    {REDIS_NODE_PFAIL,     "fail?,"},
+    {REDIS_NODE_FAIL,      "fail,"},
+    {REDIS_NODE_HANDSHAKE, "handshake,"},
+    {REDIS_NODE_NOADDR,    "noaddr,"}
+};
+
+/* Concatenate the comma separated list of node flags to the given SDS
+ * string 'ci'. */
+sds representRedisNodeFlags(sds ci, uint16_t flags) {
+    if (flags == 0) {
+        ci = sdscat(ci,"noflags,");
+    } else {
+        int i, size = sizeof(redisNodeFlagsTable)/sizeof(struct redisNodeFlags);
+        for (i = 0; i < size; i++) {
+            struct redisNodeFlags *nodeflag = redisNodeFlagsTable + i;
+            if (flags & nodeflag->flag) ci = sdscat(ci, nodeflag->name);
+        }
+    }
+    sdsIncrLen(ci,-1); /* Remove trailing comma. */
+    return ci;
+}
+
 /* Generate a csv-alike representation of the specified cluster node.
  * See clusterGenNodesDescription() top comment for more information.
  *
@@ -3344,21 +3473,13 @@ sds clusterGenNodeDescription(clusterNode *node) {
         node->port);
 
     /* Flags */
-    if (node->flags == 0) ci = sdscat(ci,"noflags,");
-    if (node->flags & REDIS_NODE_MYSELF) ci = sdscat(ci,"myself,");
-    if (node->flags & REDIS_NODE_MASTER) ci = sdscat(ci,"master,");
-    if (node->flags & REDIS_NODE_SLAVE) ci = sdscat(ci,"slave,");
-    if (node->flags & REDIS_NODE_PFAIL) ci = sdscat(ci,"fail?,");
-    if (node->flags & REDIS_NODE_FAIL) ci = sdscat(ci,"fail,");
-    if (node->flags & REDIS_NODE_HANDSHAKE) ci =sdscat(ci,"handshake,");
-    if (node->flags & REDIS_NODE_NOADDR) ci = sdscat(ci,"noaddr,");
-    if (ci[sdslen(ci)-1] == ',') ci[sdslen(ci)-1] = ' ';
+    ci = representRedisNodeFlags(ci, node->flags);
 
     /* Slave of... or just "-" */
     if (node->slaveof)
-        ci = sdscatprintf(ci,"%.40s ",node->slaveof->name);
+        ci = sdscatprintf(ci," %.40s ",node->slaveof->name);
     else
-        ci = sdscatprintf(ci,"- ");
+        ci = sdscatlen(ci," - ",3);
 
     /* Latency from the POV of this node, link status */
     ci = sdscatprintf(ci,"%lld %lld %llu %s",
@@ -3436,6 +3557,10 @@ sds clusterGenNodesDescription(int filter) {
     return ci;
 }
 
+/* -----------------------------------------------------------------------------
+ * CLUSTER command
+ * -------------------------------------------------------------------------- */
+
 int getSlotOrReply(redisClient *c, robj *o) {
     long long slot;
 
@@ -3571,7 +3696,7 @@ void clusterCommand(redisClient *c) {
         int del = !strcasecmp(c->argv[1]->ptr,"delslots");
 
         memset(slots,0,REDIS_CLUSTER_SLOTS);
-        /* Check that all the arguments are parsable and that all the
+        /* Check that all the arguments are parseable and that all the
          * slots are not already busy. */
         for (j = 2; j < c->argc; j++) {
             if ((slot = getSlotOrReply(c,c->argv[j])) == -1) {
@@ -3952,7 +4077,7 @@ void clusterCommand(redisClient *c) {
                 "configEpoch set to %llu via CLUSTER SET-CONFIG-EPOCH",
                 (unsigned long long) myself->configEpoch);
 
-            if (server.cluster->currentEpoch < epoch)
+            if (server.cluster->currentEpoch < (uint64_t)epoch)
                 server.cluster->currentEpoch = epoch;
             /* No need to fsync the config here since in the unlucky event
              * of a failure to persist the config, the conflict resolution code
@@ -4138,14 +4263,14 @@ void restoreCommand(redisClient *c) {
  * This sockets are closed when the max number we cache is reached, and also
  * in serverCron() when they are around for more than a few seconds. */
 #define MIGRATE_SOCKET_CACHE_ITEMS 64 /* max num of items in the cache. */
-#define MIGRATE_SOCKET_CACHE_TTL 10 /* close cached socekts after 10 sec. */
+#define MIGRATE_SOCKET_CACHE_TTL 10 /* close cached sockets after 10 sec. */
 
 typedef struct migrateCachedSocket {
     int fd;
     time_t last_use_time;
 } migrateCachedSocket;
 
-/* Return a TCP scoket connected with the target instance, possibly returning
+/* Return a TCP socket connected with the target instance, possibly returning
  * a cached one.
  *
  * This function is responsible of sending errors to the client if a
@@ -4154,7 +4279,7 @@ typedef struct migrateCachedSocket {
  * attempt to free it after usage.
  *
  * If the caller detects an error while using the socket, migrateCloseSocket()
- * should be called so that the connection will be craeted from scratch
+ * should be called so that the connection will be created from scratch
  * the next time. */
 int migrateGetSocket(redisClient *c, robj *host, robj *port, long timeout) {
     int fd;
@@ -4183,8 +4308,8 @@ int migrateGetSocket(redisClient *c, robj *host, robj *port, long timeout) {
     }
 
     /* Create the socket */
-    fd = anetTcpNonBlockConnect(server.neterr,c->argv[1]->ptr,
-                atoi(c->argv[2]->ptr));
+    fd = anetTcpNonBlockBindConnect(server.neterr,c->argv[1]->ptr,
+                atoi(c->argv[2]->ptr),REDIS_BIND_ADDR);
     if (fd == -1) {
         sdsfree(name);
         addReplyErrorFormat(c,"Can't connect to target node: %s",
@@ -4410,7 +4535,7 @@ void askingCommand(redisClient *c) {
     addReply(c,shared.ok);
 }
 
-/* The READONLY command is uesd by clients to enter the read-only mode.
+/* The READONLY command is used by clients to enter the read-only mode.
  * In this mode slaves will not redirect clients as long as clients access
  * with read-only commands to keys that are served by the slave's master. */
 void readonlyCommand(redisClient *c) {

src/cluster.h

@@ -11,10 +11,11 @@
 #define REDIS_CLUSTER_NAMELEN 40    /* sha1 hex length */
 #define REDIS_CLUSTER_PORT_INCR 10000 /* Cluster port = baseport + PORT_INCR */
 
-/* The following defines are amunt of time, sometimes expressed as
+/* The following defines are amount of time, sometimes expressed as
  * multiplicators of the node timeout value (when ending with MULT). */
 #define REDIS_CLUSTER_DEFAULT_NODE_TIMEOUT 15000
 #define REDIS_CLUSTER_DEFAULT_SLAVE_VALIDITY 10 /* Slave max data age factor. */
+#define REDIS_CLUSTER_DEFAULT_REQUIRE_FULL_COVERAGE 1
 #define REDIS_CLUSTER_FAIL_REPORT_VALIDITY_MULT 2 /* Fail report validity. */
 #define REDIS_CLUSTER_FAIL_UNDO_TIME_MULT 2 /* Undo fail if master is back. */
 #define REDIS_CLUSTER_FAIL_UNDO_TIME_ADD 10 /* Some additional time. */
@@ -50,7 +51,7 @@ typedef struct clusterLink {
 #define REDIS_NODE_HANDSHAKE 32 /* We have still to exchange the first ping */
 #define REDIS_NODE_NOADDR   64  /* We don't know the address of this node */
 #define REDIS_NODE_MEET 128     /* Send a MEET message to this node */
-#define REDIS_NODE_PROMOTED 256 /* Master was a slave propoted by failover */
+#define REDIS_NODE_PROMOTED 256 /* Master was a slave promoted by failover */
 #define REDIS_NODE_NULL_NAME "\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000"
 
 #define nodeIsMaster(n) ((n)->flags & REDIS_NODE_MASTER)
@@ -61,13 +62,21 @@ typedef struct clusterLink {
 #define nodeTimedOut(n) ((n)->flags & REDIS_NODE_PFAIL)
 #define nodeFailed(n) ((n)->flags & REDIS_NODE_FAIL)
 
+/* Reasons why a slave is not able to failover. */
+#define REDIS_CLUSTER_CANT_FAILOVER_NONE 0
+#define REDIS_CLUSTER_CANT_FAILOVER_DATA_AGE 1
+#define REDIS_CLUSTER_CANT_FAILOVER_WAITING_DELAY 2
+#define REDIS_CLUSTER_CANT_FAILOVER_EXPIRED 3
+#define REDIS_CLUSTER_CANT_FAILOVER_WAITING_VOTES 4
+#define REDIS_CLUSTER_CANT_FAILOVER_RELOG_PERIOD (60*5) /* seconds. */
+
 /* This structure represent elements of node->fail_reports. */
-struct clusterNodeFailReport {
+typedef struct clusterNodeFailReport {
     struct clusterNode *node;  /* Node reporting the failure condition. */
     mstime_t time;             /* Time of the last report from this node. */
-} typedef clusterNodeFailReport;
+} clusterNodeFailReport;
 
-struct clusterNode {
+typedef struct clusterNode {
     mstime_t ctime; /* Node object creation time. */
     char name[REDIS_CLUSTER_NAMELEN]; /* Node name, hex string, sha1-size */
     int flags;      /* REDIS_NODE_... */
@@ -87,8 +96,7 @@ struct clusterNode {
     int port;                   /* Latest known port of this node */
     clusterLink *link;          /* TCP/IP link with this node */
     list *fail_reports;         /* List of nodes signaling this as failing */
-};
-typedef struct clusterNode clusterNode;
+} clusterNode;
 
 typedef struct clusterState {
     clusterNode *myself;  /* This node */
@@ -107,6 +115,8 @@ typedef struct clusterState {
     int failover_auth_sent;     /* True if we already asked for votes. */
     int failover_auth_rank;     /* This slave rank for current auth request. */
     uint64_t failover_auth_epoch; /* Epoch of the current election. */
+    int cant_failover_reason;   /* Why a slave is currently not able to
+                                   failover. See the CANT_FAILOVER_* macros. */
     /* Manual failover state in common. */
     mstime_t mf_end;            /* Manual failover time limit (ms unixtime).
                                    It is zero if there is no MF in progress. */
@@ -117,7 +127,7 @@ typedef struct clusterState {
                                    or zero if stil not received. */
     int mf_can_start;           /* If non-zero signal that the manual failover
                                    can start requesting masters vote. */
-    /* The followign fields are uesd by masters to take state on elections. */
+    /* The followign fields are used by masters to take state on elections. */
     uint64_t lastVoteEpoch;     /* Epoch of the last vote granted. */
     int todo_before_sleep; /* Things to do in clusterBeforeSleep(). */
     long long stats_bus_messages_sent;  /* Num of msg sent via cluster bus. */
@@ -166,7 +176,10 @@ typedef struct {
 typedef struct {
     uint32_t channel_len;
     uint32_t message_len;
-    unsigned char bulk_data[8]; /* defined as 8 just for alignment concerns. */
+    /* We can't reclare bulk_data as bulk_data[] since this structure is
+     * nested. The 8 bytes are removed from the count during the message
+     * length computation. */
+    unsigned char bulk_data[8];
 } clusterMsgDataPublish;
 
 typedef struct {

src/config.c

@@ -73,7 +73,7 @@ void appendServerSaveParams(time_t seconds, int changes) {
     server.saveparamslen++;
 }
 
-void resetServerSaveParams() {
+void resetServerSaveParams(void) {
     zfree(server.saveparams);
     server.saveparams = NULL;
     server.saveparamslen = 0;
@@ -270,6 +270,16 @@ void loadServerConfigFromString(char *config) {
             if ((server.repl_disable_tcp_nodelay = yesnotoi(argv[1])) == -1) {
                 err = "argument must be 'yes' or 'no'"; goto loaderr;
             }
+        } else if (!strcasecmp(argv[0],"repl-diskless-sync") && argc==2) {
+            if ((server.repl_diskless_sync = yesnotoi(argv[1])) == -1) {
+                err = "argument must be 'yes' or 'no'"; goto loaderr;
+            }
+        } else if (!strcasecmp(argv[0],"repl-diskless-sync-delay") && argc==2) {
+            server.repl_diskless_sync_delay = atoi(argv[1]);
+            if (server.repl_diskless_sync_delay < 0) {
+                err = "repl-diskless-sync-delay can't be negative";
+                goto loaderr;
+            }
         } else if (!strcasecmp(argv[0],"repl-backlog-size") && argc == 2) {
             long long size = memtoll(argv[1],NULL);
             if (size <= 0) {
@@ -284,7 +294,7 @@ void loadServerConfigFromString(char *config) {
                 goto loaderr;
             }
         } else if (!strcasecmp(argv[0],"masterauth") && argc == 2) {
-        	server.masterauth = zstrdup(argv[1]);
+            server.masterauth = zstrdup(argv[1]);
         } else if (!strcasecmp(argv[0],"slave-serve-stale-data") && argc == 2) {
             if ((server.repl_serve_stale_data = yesnotoi(argv[1])) == -1) {
                 err = "argument must be 'yes' or 'no'"; goto loaderr;
@@ -358,7 +368,12 @@ void loadServerConfigFromString(char *config) {
         } else if (!strcasecmp(argv[0],"aof-rewrite-incremental-fsync") &&
                    argc == 2)
         {
-            if ((server.aof_rewrite_incremental_fsync = yesnotoi(argv[1])) == -1) {
+            if ((server.aof_rewrite_incremental_fsync =
+                 yesnotoi(argv[1])) == -1) {
+                err = "argument must be 'yes' or 'no'"; goto loaderr;
+            }
+        } else if (!strcasecmp(argv[0],"aof-load-truncated") && argc == 2) {
+            if ((server.aof_load_truncated = yesnotoi(argv[1])) == -1) {
                 err = "argument must be 'yes' or 'no'"; goto loaderr;
             }
         } else if (!strcasecmp(argv[0],"requirepass") && argc == 2) {
@@ -424,6 +439,13 @@ void loadServerConfigFromString(char *config) {
         } else if (!strcasecmp(argv[0],"cluster-config-file") && argc == 2) {
             zfree(server.cluster_configfile);
             server.cluster_configfile = zstrdup(argv[1]);
+        } else if (!strcasecmp(argv[0],"cluster-require-full-coverage") &&
+                    argc == 2)
+        {
+            if ((server.cluster_require_full_coverage = yesnotoi(argv[1])) == -1)
+            {
+                err = "argument must be 'yes' or 'no'"; goto loaderr;
+            }
         } else if (!strcasecmp(argv[0],"cluster-node-timeout") && argc == 2) {
             server.cluster_node_timeout = strtoll(argv[1],NULL,10);
             if (server.cluster_node_timeout <= 0) {
@@ -629,7 +651,7 @@ void configSetCommand(redisClient *c) {
                 server.maxclients = orig_value;
                 return;
             }
-            if (aeGetSetSize(server.el) <
+            if ((unsigned int) aeGetSetSize(server.el) <
                 server.maxclients + REDIS_EVENTLOOP_FDSET_INCR)
             {
                 if (aeResizeSetSize(server.el,
@@ -713,6 +735,11 @@ void configSetCommand(redisClient *c) {
 
         if (yn == -1) goto badfmt;
         server.aof_rewrite_incremental_fsync = yn;
+    } else if (!strcasecmp(c->argv[2]->ptr,"aof-load-truncated")) {
+        int yn = yesnotoi(o->ptr);
+
+        if (yn == -1) goto badfmt;
+        server.aof_load_truncated = yn;
     } else if (!strcasecmp(c->argv[2]->ptr,"save")) {
         int vlen, j;
         sds *v = sdssplitlen(o->ptr,sdslen(o->ptr)," ",1,&vlen);
@@ -894,6 +921,15 @@ void configSetCommand(redisClient *c) {
 
         if (yn == -1) goto badfmt;
         server.repl_disable_tcp_nodelay = yn;
+    } else if (!strcasecmp(c->argv[2]->ptr,"repl-diskless-sync")) {
+        int yn = yesnotoi(o->ptr);
+
+        if (yn == -1) goto badfmt;
+        server.repl_diskless_sync = yn;
+    } else if (!strcasecmp(c->argv[2]->ptr,"repl-diskless-sync-delay")) {
+        if (getLongLongFromObject(o,&ll) == REDIS_ERR ||
+            ll < 0) goto badfmt;
+        server.repl_diskless_sync_delay = ll;
     } else if (!strcasecmp(c->argv[2]->ptr,"slave-priority")) {
         if (getLongLongFromObject(o,&ll) == REDIS_ERR ||
             ll < 0) goto badfmt;
@@ -908,6 +944,11 @@ void configSetCommand(redisClient *c) {
             ll < 0) goto badfmt;
         server.repl_min_slaves_max_lag = ll;
         refreshGoodSlavesCount();
+    } else if (!strcasecmp(c->argv[2]->ptr,"cluster-require-full-coverage")) {
+        int yn = yesnotoi(o->ptr);
+
+        if (yn == -1) goto badfmt;
+        server.cluster_require_full_coverage = yn;
     } else if (!strcasecmp(c->argv[2]->ptr,"cluster-node-timeout")) {
         if (getLongLongFromObject(o,&ll) == REDIS_ERR ||
             ll <= 0) goto badfmt;
@@ -1027,8 +1068,11 @@ void configGetCommand(redisClient *c) {
     config_get_numerical_field("cluster-node-timeout",server.cluster_node_timeout);
     config_get_numerical_field("cluster-migration-barrier",server.cluster_migration_barrier);
     config_get_numerical_field("cluster-slave-validity-factor",server.cluster_slave_validity_factor);
+    config_get_numerical_field("repl-diskless-sync-delay",server.repl_diskless_sync_delay);
 
     /* Bool (yes/no) values */
+    config_get_bool_field("cluster-require-full-coverage",
+            server.cluster_require_full_coverage);
     config_get_bool_field("no-appendfsync-on-rewrite",
             server.aof_no_fsync_on_rewrite);
     config_get_bool_field("slave-serve-stale-data",
@@ -1043,8 +1087,12 @@ void configGetCommand(redisClient *c) {
     config_get_bool_field("activerehashing", server.activerehashing);
     config_get_bool_field("repl-disable-tcp-nodelay",
             server.repl_disable_tcp_nodelay);
+    config_get_bool_field("repl-diskless-sync",
+            server.repl_diskless_sync);
     config_get_bool_field("aof-rewrite-incremental-fsync",
             server.aof_rewrite_incremental_fsync);
+    config_get_bool_field("aof-load-truncated",
+            server.aof_load_truncated);
 
     /* Everything we can't handle with macros follows. */
 
@@ -1415,7 +1463,7 @@ void rewriteConfigStringOption(struct rewriteConfigState *state, char *option, c
         return;
     }
 
-    /* Compare the strings as sds strings to have a binary safe comparison. */
+    /* Set force to zero if the value is set to its default. */
     if (defvalue && strcmp(value,defvalue) == 0) force = 0;
 
     line = sdsnew(option);
@@ -1467,7 +1515,7 @@ void rewriteConfigEnumOption(struct rewriteConfigState *state, char *option, int
     rewriteConfigRewriteLine(state,option,line,force);
 }
 
-/* Rewrite the syslog-fability option. */
+/* Rewrite the syslog-facility option. */
 void rewriteConfigSyslogfacilityOption(struct rewriteConfigState *state) {
     int value = server.syslog_facility, j;
     int force = value != LOG_LOCAL0;
@@ -1766,6 +1814,8 @@ int rewriteConfig(char *path) {
     rewriteConfigBytesOption(state,"repl-backlog-size",server.repl_backlog_size,REDIS_DEFAULT_REPL_BACKLOG_SIZE);
     rewriteConfigBytesOption(state,"repl-backlog-ttl",server.repl_backlog_time_limit,REDIS_DEFAULT_REPL_BACKLOG_TIME_LIMIT);
     rewriteConfigYesNoOption(state,"repl-disable-tcp-nodelay",server.repl_disable_tcp_nodelay,REDIS_DEFAULT_REPL_DISABLE_TCP_NODELAY);
+    rewriteConfigYesNoOption(state,"repl-diskless-sync",server.repl_diskless_sync,REDIS_DEFAULT_REPL_DISKLESS_SYNC);
+    rewriteConfigNumericalOption(state,"repl-diskless-sync-delay",server.repl_diskless_sync_delay,REDIS_DEFAULT_REPL_DISKLESS_SYNC_DELAY);
     rewriteConfigNumericalOption(state,"slave-priority",server.slave_priority,REDIS_DEFAULT_SLAVE_PRIORITY);
     rewriteConfigNumericalOption(state,"min-slaves-to-write",server.repl_min_slaves_to_write,REDIS_DEFAULT_MIN_SLAVES_TO_WRITE);
     rewriteConfigNumericalOption(state,"min-slaves-max-lag",server.repl_min_slaves_max_lag,REDIS_DEFAULT_MIN_SLAVES_MAX_LAG);
@@ -1794,6 +1844,7 @@ int rewriteConfig(char *path) {
     rewriteConfigNumericalOption(state,"lua-time-limit",server.lua_time_limit,REDIS_LUA_TIME_LIMIT);
     rewriteConfigYesNoOption(state,"cluster-enabled",server.cluster_enabled,0);
     rewriteConfigStringOption(state,"cluster-config-file",server.cluster_configfile,REDIS_DEFAULT_CLUSTER_CONFIG_FILE);
+    rewriteConfigYesNoOption(state,"cluster-require-full-coverage",server.cluster_require_full_coverage,REDIS_CLUSTER_DEFAULT_REQUIRE_FULL_COVERAGE);
     rewriteConfigNumericalOption(state,"cluster-node-timeout",server.cluster_node_timeout,REDIS_CLUSTER_DEFAULT_NODE_TIMEOUT);
     rewriteConfigNumericalOption(state,"cluster-migration-barrier",server.cluster_migration_barrier,REDIS_CLUSTER_DEFAULT_MIGRATION_BARRIER);
     rewriteConfigNumericalOption(state,"cluster-slave-validity-factor",server.cluster_slave_validity_factor,REDIS_CLUSTER_DEFAULT_SLAVE_VALIDITY);
@@ -1813,6 +1864,7 @@ int rewriteConfig(char *path) {
     rewriteConfigClientoutputbufferlimitOption(state);
     rewriteConfigNumericalOption(state,"hz",server.hz,REDIS_DEFAULT_HZ);
     rewriteConfigYesNoOption(state,"aof-rewrite-incremental-fsync",server.aof_rewrite_incremental_fsync,REDIS_DEFAULT_AOF_REWRITE_INCREMENTAL_FSYNC);
+    rewriteConfigYesNoOption(state,"aof-load-truncated",server.aof_load_truncated,REDIS_DEFAULT_AOF_LOAD_TRUNCATED);
     if (server.sentinel_mode) rewriteConfigSentinelOption(state);
 
     /* Step 3: remove all the orphaned lines in the old file, that is, lines

src/config.h

@@ -187,9 +187,14 @@ void setproctitle(const char *fmt, ...);
 
 #if (__i386 || __amd64 || __powerpc__) && __GNUC__
 #define GNUC_VERSION (__GNUC__ * 10000 + __GNUC_MINOR__ * 100 + __GNUC_PATCHLEVEL__)
-#if (GNUC_VERSION >= 40100) || defined(__clang__)
+#if defined(__clang__)
 #define HAVE_ATOMIC
 #endif
+#if (defined(__GLIBC__) && defined(__GLIBC_PREREQ))
+#if (GNUC_VERSION >= 40100 && __GLIBC_PREREQ(2, 6))
+#define HAVE_ATOMIC
+#endif
+#endif
 #endif
 
 #endif

src/db.c

@@ -381,7 +381,7 @@ void scanCallback(void *privdata, const dictEntry *de) {
     } else if (o->type == REDIS_ZSET) {
         key = dictGetKey(de);
         incrRefCount(key);
-        val = createStringObjectFromLongDouble(*(double*)dictGetVal(de));
+        val = createStringObjectFromLongDouble(*(double*)dictGetVal(de),0);
     } else {
         redisPanic("Type not handled in SCAN callback.");
     }
@@ -421,9 +421,7 @@ int parseScanCursorOrReply(redisClient *c, robj *o, unsigned long *cursor) {
  * In the case of a Hash object the function returns both the field and value
  * of every element on the Hash. */
 void scanGenericCommand(redisClient *c, robj *o, unsigned long cursor) {
-    int rv;
     int i, j;
-    char buf[REDIS_LONGSTR_SIZE];
     list *keys = listCreate();
     listNode *node, *nextnode;
     long count = 10;
@@ -495,6 +493,11 @@ void scanGenericCommand(redisClient *c, robj *o, unsigned long cursor) {
 
     if (ht) {
         void *privdata[2];
+        /* We set the max number of iterations to ten times the specified
+         * COUNT, so if the hash table is in a pathological state (very
+         * sparsely populated) we avoid to block too much time at the cost
+         * of returning no or very few elements. */
+        long maxiterations = count*10;
 
         /* We pass two pointers to the callback: the list to which it will
          * add new elements, and the object containing the dictionary so that
@@ -503,7 +506,9 @@ void scanGenericCommand(redisClient *c, robj *o, unsigned long cursor) {
         privdata[1] = o;
         do {
             cursor = dictScan(ht, cursor, scanCallback, privdata);
-        } while (cursor && listLength(keys) < count);
+        } while (cursor &&
+              maxiterations-- &&
+              listLength(keys) < (unsigned long)count);
     } else if (o->type == REDIS_SET) {
         int pos = 0;
         int64_t ll;
@@ -577,9 +582,7 @@ void scanGenericCommand(redisClient *c, robj *o, unsigned long cursor) {
 
     /* Step 4: Reply to the client. */
     addReplyMultiBulkLen(c, 2);
-    rv = snprintf(buf, sizeof(buf), "%lu", cursor);
-    redisAssert(rv < sizeof(buf));
-    addReplyBulkCBuffer(c, buf, rv);
+    addReplyBulkLongLong(c,cursor);
 
     addReplyMultiBulkLen(c, listLength(keys));
     while ((node = listFirst(keys)) != NULL) {
@@ -707,6 +710,7 @@ void moveCommand(redisClient *c) {
     robj *o;
     redisDb *src, *dst;
     int srcid;
+    long long dbid;
 
     if (server.cluster_enabled) {
         addReplyError(c,"MOVE is not allowed in cluster mode");
@@ -716,7 +720,11 @@ void moveCommand(redisClient *c) {
     /* Obtain source and target DB pointers */
     src = c->db;
     srcid = c->db->id;
-    if (selectDb(c,atoi(c->argv[2]->ptr)) == REDIS_ERR) {
+
+    if (getLongLongFromObject(c->argv[2],&dbid) == REDIS_ERR ||
+        dbid < INT_MIN || dbid > INT_MAX ||
+        selectDb(c,dbid) == REDIS_ERR)
+    {
         addReply(c,shared.outofrangeerr);
         return;
     }
@@ -1076,7 +1084,7 @@ int *evalGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys)
  * follow in SQL-alike style. Here we parse just the minimum in order to
  * correctly identify keys in the "STORE" option. */
 int *sortGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys) {
-    int i, j, num, *keys;
+    int i, j, num, *keys, found_store = 0;
     REDIS_NOTUSED(cmd);
 
     num = 0;
@@ -1107,12 +1115,13 @@ int *sortGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys)
                 /* Note: we don't increment "num" here and continue the loop
                  * to be sure to process the *last* "STORE" option if multiple
                  * ones are provided. This is same behavior as SORT. */
+                found_store = 1;
                 keys[num] = i+1; /* <store-key> */
                 break;
             }
         }
     }
-    *numkeys = num;
+    *numkeys = num + found_store;
     return keys;
 }
 

src/debug.c

@@ -301,7 +301,7 @@ void debugCommand(redisClient *c) {
             "lru:%d lru_seconds_idle:%llu",
             (void*)val, val->refcount,
             strenc, (long long) rdbSavedObjectLen(val),
-            val->lru, estimateObjectIdleTime(val));
+            val->lru, estimateObjectIdleTime(val)/1000);
     } else if (!strcasecmp(c->argv[1]->ptr,"sdslen") && c->argc == 3) {
         dictEntry *de;
         robj *val;
@@ -325,7 +325,8 @@ void debugCommand(redisClient *c) {
                 (long long) sdslen(val->ptr),
                 (long long) sdsavail(val->ptr));
         }
-    } else if (!strcasecmp(c->argv[1]->ptr,"populate") && c->argc == 3) {
+    } else if (!strcasecmp(c->argv[1]->ptr,"populate") &&
+               (c->argc == 3 || c->argc == 4)) {
         long keys, j;
         robj *key, *val;
         char buf[128];
@@ -334,7 +335,8 @@ void debugCommand(redisClient *c) {
             return;
         dictExpand(c->db->dict,keys);
         for (j = 0; j < keys; j++) {
-            snprintf(buf,sizeof(buf),"key:%lu",j);
+            snprintf(buf,sizeof(buf),"%s:%lu",
+                (c->argc == 3) ? "key" : (char*)c->argv[3]->ptr, j);
             key = createStringObject(buf,strlen(buf));
             if (lookupKeyRead(c->db,key) != NULL) {
                 decrRefCount(key);
@@ -850,7 +852,7 @@ void sigsegvHandler(int sig, siginfo_t *info, void *secret) {
 
     redisLog(REDIS_WARNING,
 "\n=== REDIS BUG REPORT END. Make sure to include from START to END. ===\n\n"
-"       Please report the crash opening an issue on github:\n\n"
+"       Please report the crash by opening an issue on github:\n\n"
 "           http://github.com/antirez/redis/issues\n\n"
 "  Suspect RAM error? Use redis-server --test-memory to verify it.\n\n"
 );

src/dict.c

@@ -79,12 +79,6 @@ unsigned int dictIntHashFunction(unsigned int key)
     return key;
 }
 
-/* Identity hash function for integer keys */
-unsigned int dictIdentityHashFunction(unsigned int key)
-{
-    return key;
-}
-
 static uint32_t dict_hash_function_seed = 5381;
 
 void dictSetHashFunctionSeed(uint32_t seed) {
@@ -257,7 +251,7 @@ int dictRehash(dict *d, int n) {
 
         /* Note that rehashidx can't overflow as we are sure there are more
          * elements because ht[0].used != 0 */
-        assert(d->ht[0].size > (unsigned)d->rehashidx);
+        assert(d->ht[0].size > (unsigned long)d->rehashidx);
         while(d->ht[0].table[d->rehashidx] == NULL) d->rehashidx++;
         de = d->ht[0].table[d->rehashidx];
         /* Move all the keys in this bucket from the old to the new hash HT */
@@ -576,7 +570,7 @@ dictEntry *dictNext(dictIterator *iter)
                     iter->fingerprint = dictFingerprint(iter->d);
             }
             iter->index++;
-            if (iter->index >= (signed) ht->size) {
+            if (iter->index >= (long) ht->size) {
                 if (dictIsRehashing(iter->d) && iter->table == 0) {
                     iter->table++;
                     iter->index = 0;
@@ -668,9 +662,9 @@ dictEntry *dictGetRandomKey(dict *d)
  * statistics. However the function is much faster than dictGetRandomKey()
  * at producing N elements, and the elements are guaranteed to be non
  * repeating. */
-int dictGetRandomKeys(dict *d, dictEntry **des, int count) {
+unsigned int dictGetRandomKeys(dict *d, dictEntry **des, unsigned int count) {
     int j; /* internal hash table id, 0 or 1. */
-    int stored = 0;
+    unsigned int stored = 0;
 
     if (dictSize(d) < count) count = dictSize(d);
     while(stored < count) {
@@ -715,72 +709,72 @@ static unsigned long rev(unsigned long v) {
 
 /* dictScan() is used to iterate over the elements of a dictionary.
  *
- * Iterating works in the following way:
+ * Iterating works the following way:
  *
  * 1) Initially you call the function using a cursor (v) value of 0.
  * 2) The function performs one step of the iteration, and returns the
- *    new cursor value that you must use in the next call.
+ *    new cursor value you must use in the next call.
  * 3) When the returned cursor is 0, the iteration is complete.
  *
- * The function guarantees that all the elements that are present in the
- * dictionary from the start to the end of the iteration are returned.
- * However it is possible that some element is returned multiple time.
+ * The function guarantees all elements present in the
+ * dictionary get returned between the start and end of the iteration.
+ * However it is possible some elements get returned multiple times.
  *
- * For every element returned, the callback 'fn' passed as argument is
- * called, with 'privdata' as first argument and the dictionar entry
+ * For every element returned, the callback argument 'fn' is
+ * called with 'privdata' as first argument and the dictionary entry
  * 'de' as second argument.
  *
  * HOW IT WORKS.
  *
- * The algorithm used in the iteration was designed by Pieter Noordhuis.
+ * The iteration algorithm was designed by Pieter Noordhuis.
  * The main idea is to increment a cursor starting from the higher order
- * bits, that is, instead of incrementing the cursor normally, the bits
+ * bits. That is, instead of incrementing the cursor normally, the bits
  * of the cursor are reversed, then the cursor is incremented, and finally
  * the bits are reversed again.
  *
- * This strategy is needed because the hash table may be resized from one
- * call to the other call of the same iteration.
+ * This strategy is needed because the hash table may be resized between
+ * iteration calls.
  *
  * dict.c hash tables are always power of two in size, and they
  * use chaining, so the position of an element in a given table is given
- * always by computing the bitwise AND between Hash(key) and SIZE-1
+ * by computing the bitwise AND between Hash(key) and SIZE-1
  * (where SIZE-1 is always the mask that is equivalent to taking the rest
  *  of the division between the Hash of the key and SIZE).
  *
  * For example if the current hash table size is 16, the mask is
- * (in binary) 1111. The position of a key in the hash table will be always
+ * (in binary) 1111. The position of a key in the hash table will always be
  * the last four bits of the hash output, and so forth.
  *
  * WHAT HAPPENS IF THE TABLE CHANGES IN SIZE?
  *
- * If the hash table grows, elements can go anyway in one multiple of
- * the old bucket: for example let's say that we already iterated with
- * a 4 bit cursor 1100, since the mask is 1111 (hash table size = 16).
+ * If the hash table grows, elements can go anywhere in one multiple of
+ * the old bucket: for example let's say we already iterated with
+ * a 4 bit cursor 1100 (the mask is 1111 because hash table size = 16).
  *
- * If the hash table will be resized to 64 elements, and the new mask will
- * be 111111, the new buckets that you obtain substituting in ??1100
- * either 0 or 1, can be targeted only by keys that we already visited
+ * If the hash table will be resized to 64 elements, then the new mask will
+ * be 111111. The new buckets you obtain by substituting in ??1100
+ * with either 0 or 1 can be targeted only by keys we already visited
  * when scanning the bucket 1100 in the smaller hash table.
  *
  * By iterating the higher bits first, because of the inverted counter, the
- * cursor does not need to restart if the table size gets bigger, and will
- * just continue iterating with cursors that don't have '1100' at the end,
- * nor any other combination of final 4 bits already explored.
+ * cursor does not need to restart if the table size gets bigger. It will
+ * continue iterating using cursors without '1100' at the end, and also
+ * without any other combination of the final 4 bits already explored.
  *
  * Similarly when the table size shrinks over time, for example going from
- * 16 to 8, If a combination of the lower three bits (the mask for size 8
- * is 111) was already completely explored, it will not be visited again
- * as we are sure that, we tried for example, both 0111 and 1111 (all the
+ * 16 to 8, if a combination of the lower three bits (the mask for size 8
+ * is 111) were already completely explored, it would not be visited again
+ * because we are sure we tried, for example, both 0111 and 1111 (all the
  * variations of the higher bit) so we don't need to test it again.
  *
  * WAIT... YOU HAVE *TWO* TABLES DURING REHASHING!
  *
- * Yes, this is true, but we always iterate the smaller one of the tables,
- * testing also all the expansions of the current cursor into the larger
- * table. So for example if the current cursor is 101 and we also have a
+ * Yes, this is true, but we always iterate the smaller table first, then
+ * we test all the expansions of the current cursor into the larger
+ * table. For example if the current cursor is 101 and we also have a
  * larger table of size 16, we also test (0)101 and (1)101 inside the larger
  * table. This reduces the problem back to having only one table, where
- * the larger one, if exists, is just an expansion of the smaller one.
+ * the larger one, if it exists, is just an expansion of the smaller one.
  *
  * LIMITATIONS
  *
@@ -789,11 +783,11 @@ static unsigned long rev(unsigned long v) {
  *
  * The disadvantages resulting from this design are:
  *
- * 1) It is possible that we return duplicated elements. However this is usually
+ * 1) It is possible we return elements more than once. However this is usually
  *    easy to deal with in the application level.
  * 2) The iterator must return multiple elements per call, as it needs to always
  *    return all the keys chained in a given bucket, and all the expansions, so
- *    we are sure we don't miss keys moving.
+ *    we are sure we don't miss keys moving during rehashing.
  * 3) The reverse cursor is somewhat hard to understand at first, but this
  *    comment is supposed to help.
  */

src/dict.h

@@ -50,6 +50,7 @@ typedef struct dictEntry {
         void *val;
         uint64_t u64;
         int64_t s64;
+        double d;
     } v;
     struct dictEntry *next;
 } dictEntry;
@@ -76,7 +77,7 @@ typedef struct dict {
     dictType *type;
     void *privdata;
     dictht ht[2];
-    int rehashidx; /* rehashing not in progress if rehashidx == -1 */
+    long rehashidx; /* rehashing not in progress if rehashidx == -1 */
     int iterators; /* number of iterators currently running */
 } dict;
 
@@ -86,9 +87,11 @@ typedef struct dict {
  * should be called while iterating. */
 typedef struct dictIterator {
     dict *d;
-    int table, index, safe;
+    long index;
+    int table, safe;
     dictEntry *entry, *nextEntry;
-    long long fingerprint; /* unsafe iterator fingerprint for misuse detection */
+    /* unsafe iterator fingerprint for misuse detection. */
+    long long fingerprint;
 } dictIterator;
 
 typedef void (dictScanFunction)(void *privdata, const dictEntry *de);
@@ -114,6 +117,9 @@ typedef void (dictScanFunction)(void *privdata, const dictEntry *de);
 #define dictSetUnsignedIntegerVal(entry, _val_) \
     do { entry->v.u64 = _val_; } while(0)
 
+#define dictSetDoubleVal(entry, _val_) \
+    do { entry->v.d = _val_; } while(0)
+
 #define dictFreeKey(d, entry) \
     if ((d)->type->keyDestructor) \
         (d)->type->keyDestructor((d)->privdata, (entry)->key)
@@ -135,9 +141,10 @@ typedef void (dictScanFunction)(void *privdata, const dictEntry *de);
 #define dictGetVal(he) ((he)->v.val)
 #define dictGetSignedIntegerVal(he) ((he)->v.s64)
 #define dictGetUnsignedIntegerVal(he) ((he)->v.u64)
+#define dictGetDoubleVal(he) ((he)->v.d)
 #define dictSlots(d) ((d)->ht[0].size+(d)->ht[1].size)
 #define dictSize(d) ((d)->ht[0].used+(d)->ht[1].used)
-#define dictIsRehashing(ht) ((ht)->rehashidx != -1)
+#define dictIsRehashing(d) ((d)->rehashidx != -1)
 
 /* API */
 dict *dictCreate(dictType *type, void *privDataPtr);
@@ -157,7 +164,7 @@ dictIterator *dictGetSafeIterator(dict *d);
 dictEntry *dictNext(dictIterator *iter);
 void dictReleaseIterator(dictIterator *iter);
 dictEntry *dictGetRandomKey(dict *d);
-int dictGetRandomKeys(dict *d, dictEntry **des, int count);
+unsigned int dictGetRandomKeys(dict *d, dictEntry **des, unsigned int count);
 void dictPrintStats(dict *d);
 unsigned int dictGenHashFunction(const void *key, int len);
 unsigned int dictGenCaseHashFunction(const unsigned char *buf, int len);

src/fmacros.h

@@ -36,6 +36,10 @@
 #define _GNU_SOURCE
 #endif
 
+#if defined(_AIX)
+#define _ALL_SOURCE
+#endif
+
 #if defined(__linux__) || defined(__OpenBSD__)
 #define _XOPEN_SOURCE 700
 /*

src/hyperloglog.c

@@ -1213,7 +1213,7 @@ void pfcountCommand(redisClient *c) {
         for (j = 1; j < c->argc; j++) {
             /* Check type and size. */
             robj *o = lookupKeyRead(c->db,c->argv[j]);
-            if (o == NULL) continue; /* Assume empty HLL for non existing var. */
+            if (o == NULL) continue; /* Assume empty HLL for non existing var.*/
             if (isHLLObjectOrReply(c,o) != REDIS_OK) return;
 
             /* Merge with this HLL with our 'max' HHL by setting max[i]
@@ -1349,7 +1349,7 @@ void pfmergeCommand(redisClient *c) {
  * Something that is not easy to test from within the outside. */
 #define HLL_TEST_CYCLES 1000
 void pfselftestCommand(redisClient *c) {
-    int j, i;
+    unsigned int j, i;
     sds bitcounters = sdsnewlen(NULL,HLL_DENSE_SIZE);
     struct hllhdr *hdr = (struct hllhdr*) bitcounters, *hdr2;
     robj *o = NULL;
@@ -1386,7 +1386,7 @@ void pfselftestCommand(redisClient *c) {
      * The test adds unique elements and check that the estimated value
      * is always reasonable bounds.
      *
-     * We check that the error is smaller than 4 times than the expected
+     * We check that the error is smaller than a few times than the expected
      * standard error, to make it very unlikely for the test to fail because
      * of a "bad" run.
      *
@@ -1422,8 +1422,16 @@ void pfselftestCommand(redisClient *c) {
         /* Check error. */
         if (j == checkpoint) {
             int64_t abserr = checkpoint - (int64_t)hllCount(hdr,NULL);
+            uint64_t maxerr = ceil(relerr*6*checkpoint);
+
+            /* Adjust the max error we expect for cardinality 10
+             * since from time to time it is statistically likely to get
+             * much higher error due to collision, resulting into a false
+             * positive. */
+            if (j == 10) maxerr = 1;
+
             if (abserr < 0) abserr = -abserr;
-            if (abserr > (uint64_t)(relerr*4*checkpoint)) {
+            if (abserr > (int64_t)maxerr) {
                 addReplyErrorFormat(c,
                     "TESTFAILED Too big error. card:%llu abserr:%llu",
                     (unsigned long long) checkpoint,

src/intset.c

@@ -133,7 +133,7 @@ static uint8_t intsetSearch(intset *is, int64_t value, uint32_t *pos) {
     }
 
     while(max >= min) {
-        mid = (min+max)/2;
+        mid = ((unsigned int)min + (unsigned int)max) >> 1;
         cur = _intsetGet(is,mid);
         if (value > cur) {
             min = mid+1;

src/latency.c

@@ -37,6 +37,7 @@
 
 /* Dictionary type for latency events. */
 int dictStringKeyCompare(void *privdata, const void *key1, const void *key2) {
+    REDIS_NOTUSED(privdata);
     return strcmp(key1,key2) == 0;
 }
 
@@ -55,6 +56,32 @@ dictType latencyTimeSeriesDictType = {
     dictVanillaFree             /* val destructor */
 };
 
+/* ------------------------- Utility functions ------------------------------ */
+
+#ifdef __linux__
+/* Returns 1 if Transparent Huge Pages support is enabled in the kernel.
+ * Otherwise (or if we are unable to check) 0 is returned. */
+int THPIsEnabled(void) {
+    char buf[1024];
+
+    FILE *fp = fopen("/sys/kernel/mm/transparent_hugepage/enabled","r");
+    if (!fp) return 0;
+    if (fgets(buf,sizeof(buf),fp) == NULL) {
+        fclose(fp);
+        return 0;
+    }
+    fclose(fp);
+    return (strstr(buf,"[never]") == NULL) ? 1 : 0;
+}
+#endif
+
+/* Report the amount of AnonHugePages in smap, in bytes. If the return
+ * value of the function is non-zero, the process is being targeted by
+ * THP support, and is likely to have memory usage / latency issues. */
+int THPGetAnonHugePagesSize(void) {
+    return zmalloc_get_smap_bytes_by_field("AnonHugePages:");
+}
+
 /* ---------------------------- Latency API --------------------------------- */
 
 /* Latency monitor initialization. We just need to create the dictionary
@@ -202,6 +229,7 @@ sds createLatencyReport(void) {
     int advise_hz = 0;              /* Use higher HZ. */
     int advise_large_objects = 0;   /* Deletion of large objects. */
     int advise_relax_fsync_policy = 0; /* appendfsync always is slow. */
+    int advise_disable_thp = 0;     /* AnonHugePages detected. */
     int advices = 0;
 
     /* Return ASAP if the latency engine is disabled and it looks like it
@@ -345,9 +373,15 @@ sds createLatencyReport(void) {
     }
     dictReleaseIterator(di);
 
-    if (eventnum == 0) {
+    /* Add non event based advices. */
+    if (THPGetAnonHugePagesSize() > 0) {
+        advise_disable_thp = 1;
+        advices++;
+    }
+
+    if (eventnum == 0 && advices == 0) {
         report = sdscat(report,"Dave, no latency spike was observed during the lifetime of this Redis instance, not in the slightest bit. I honestly think you ought to sit down calmly, take a stress pill, and think things over.\n");
-    } else if (advices == 0) {
+    } else if (eventnum > 0 && advices == 0) {
         report = sdscat(report,"\nWhile there are latency events logged, I'm not able to suggest any easy fix. Please use the Redis community to get some help, providing this report in your help request.\n");
     } else {
         /* Add all the suggestions accumulated so far. */
@@ -368,7 +402,7 @@ sds createLatencyReport(void) {
         }
 
         if (advise_slowlog_inspect) {
-            report = sdscat(report,"- Check your Slow Log to understand what are the commads you are running which are too slow to execute. Please check http://redis.io/commands/slowlog for more information.\n");
+            report = sdscat(report,"- Check your Slow Log to understand what are the commands you are running which are too slow to execute. Please check http://redis.io/commands/slowlog for more information.\n");
         }
 
         /* Intrinsic latency. */
@@ -417,6 +451,10 @@ sds createLatencyReport(void) {
         if (advise_large_objects) {
             report = sdscat(report,"- Deleting, expiring or evicting (because of maxmemory policy) large objects is a blocking operation. If you have very large objects that are often deleted, expired, or evicted, try to fragment those objects into multiple smaller objects.\n");
         }
+
+        if (advise_disable_thp) {
+            report = sdscat(report,"- I detected a non zero amount of anonymous huge pages used by your process. This creates very serious latency events in different conditions, especially when Redis is persisting on disk. To disable THP support use the command 'echo never > /sys/kernel/mm/transparent_hugepage/enabled', make sure to also add it into /etc/rc.local so that the command will be executed again after a reboot. Note that even if you have already disabled THP, you still need to restart the Redis process to get rid of the huge pages already created.\n");
+        }
     }
 
     return report;
@@ -524,8 +562,11 @@ void latencyCommand(redisClient *c) {
     if (!strcasecmp(c->argv[1]->ptr,"history") && c->argc == 3) {
         /* LATENCY HISTORY <event> */
         ts = dictFetchValue(server.latency_events,c->argv[2]->ptr);
-        if (ts == NULL) goto nodataerr;
-        latencyCommandReplyWithSamples(c,ts);
+        if (ts == NULL) {
+            addReplyMultiBulkLen(c,0);
+        } else {
+            latencyCommandReplyWithSamples(c,ts);
+        }
     } else if (!strcasecmp(c->argv[1]->ptr,"graph") && c->argc == 3) {
         /* LATENCY GRAPH <event> */
         sds graph;

src/latency.h

@@ -63,6 +63,7 @@ struct latencyStats {
 
 void latencyMonitorInit(void);
 void latencyAddSample(char *event, mstime_t latency);
+int THPIsEnabled(void);
 
 /* Latency monitoring macros. */
 

src/memtest.c

@@ -237,6 +237,7 @@ void memtest_test(size_t megabytes, int passes) {
         memtest_progress_end();
         memtest_compare_times(m,bytes,pass,4);
     }
+    free(m);
 }
 
 void memtest_non_destructive_invert(void *addr, size_t size) {

src/multi.c

@@ -72,7 +72,7 @@ void queueMultiCommand(redisClient *c) {
 void discardTransaction(redisClient *c) {
     freeClientMultiState(c);
     initClientMultiState(c);
-    c->flags &= ~(REDIS_MULTI|REDIS_DIRTY_CAS|REDIS_DIRTY_EXEC);;
+    c->flags &= ~(REDIS_MULTI|REDIS_DIRTY_CAS|REDIS_DIRTY_EXEC);
     unwatchAllKeys(c);
 }
 

src/networking.c

@@ -100,6 +100,7 @@ redisClient *createClient(int fd) {
     c->ctime = c->lastinteraction = server.unixtime;
     c->authenticated = 0;
     c->replstate = REDIS_REPL_NONE;
+    c->repl_put_online_on_ack = 0;
     c->reploff = 0;
     c->repl_ack_off = 0;
     c->repl_ack_time = 0;
@@ -678,12 +679,8 @@ void freeClient(redisClient *c) {
 
     /* Log link disconnection with slave */
     if ((c->flags & REDIS_SLAVE) && !(c->flags & REDIS_MONITOR)) {
-        char ip[REDIS_IP_STR_LEN];
-
-        if (anetPeerToString(c->fd,ip,sizeof(ip),NULL) != -1) {
-            redisLog(REDIS_WARNING,"Connection with slave %s:%d lost.",
-                ip, c->slave_listening_port);
-        }
+        redisLog(REDIS_WARNING,"Connection with slave %s lost.",
+            replicationGetSlaveName(c));
     }
 
     /* Free the query buffer */
@@ -929,8 +926,10 @@ int processInlineBuffer(redisClient *c) {
     sdsrange(c->querybuf,querylen+2,-1);
 
     /* Setup argv array on client structure */
-    if (c->argv) zfree(c->argv);
-    c->argv = zmalloc(sizeof(robj*)*argc);
+    if (argc) {
+        if (c->argv) zfree(c->argv);
+        c->argv = zmalloc(sizeof(robj*)*argc);
+    }
 
     /* Create redis objects for all arguments. */
     for (c->argc = 0, j = 0; j < argc; j++) {
@@ -1051,7 +1050,7 @@ int processMultibulkBuffer(redisClient *c) {
                 qblen = sdslen(c->querybuf);
                 /* Hint the sds library about the amount of bytes this string is
                  * going to contain. */
-                if (qblen < ll+2)
+                if (qblen < (size_t)ll+2)
                     c->querybuf = sdsMakeRoomFor(c->querybuf,ll+2-qblen);
             }
             c->bulklen = ll;
@@ -1230,9 +1229,9 @@ void formatPeerId(char *peerid, size_t peerid_len, char *ip, int port) {
 }
 
 /* A Redis "Peer ID" is a colon separated ip:port pair.
- * For IPv4 it's in the form x.y.z.k:pork, example: "127.0.0.1:1234".
+ * For IPv4 it's in the form x.y.z.k:port, example: "127.0.0.1:1234".
  * For IPv6 addresses we use [] around the IP part, like in "[::1]:1234".
- * For Unix socekts we use path:0, like in "/tmp/redis:0".
+ * For Unix sockets we use path:0, like in "/tmp/redis:0".
  *
  * A Peer ID always fits inside a buffer of REDIS_PEER_ID_LEN bytes, including
  * the null term.
@@ -1259,7 +1258,7 @@ int genClientPeerId(redisClient *client, char *peerid, size_t peerid_len) {
 }
 
 /* This function returns the client peer id, by creating and caching it
- * if client->perrid is NULL, otherwise returning the cached value.
+ * if client->peerid is NULL, otherwise returning the cached value.
  * The Peer ID never changes during the life of the client, however it
  * is expensive to compute. */
 char *getClientPeerId(redisClient *c) {
@@ -1561,7 +1560,7 @@ unsigned long getClientOutputBufferMemoryUsage(redisClient *c) {
 int getClientType(redisClient *c) {
     if ((c->flags & REDIS_SLAVE) && !(c->flags & REDIS_MONITOR))
         return REDIS_CLIENT_TYPE_SLAVE;
-    if (dictSize(c->pubsub_channels) || listLength(c->pubsub_patterns))
+    if (c->flags & REDIS_PUBSUB)
         return REDIS_CLIENT_TYPE_PUBSUB;
     return REDIS_CLIENT_TYPE_NORMAL;
 }

src/object.c

@@ -109,26 +109,44 @@ robj *createStringObjectFromLongLong(long long value) {
     return o;
 }
 
-/* Note: this function is defined into object.c since here it is where it
- * belongs but it is actually designed to be used just for INCRBYFLOAT */
-robj *createStringObjectFromLongDouble(long double value) {
+/* Create a string object from a long double. If humanfriendly is non-zero
+ * it does not use exponential format and trims trailing zeroes at the end,
+ * however this results in loss of precision. Otherwise exp format is used
+ * and the output of snprintf() is not modified.
+ *
+ * The 'humanfriendly' option is used for INCRBYFLOAT and HINCRBYFLOAT. */
+robj *createStringObjectFromLongDouble(long double value, int humanfriendly) {
     char buf[256];
     int len;
 
-    /* We use 17 digits precision since with 128 bit floats that precision
-     * after rounding is able to represent most small decimal numbers in a way
-     * that is "non surprising" for the user (that is, most small decimal
-     * numbers will be represented in a way that when converted back into
-     * a string are exactly the same as what the user typed.) */
-    len = snprintf(buf,sizeof(buf),"%.17Lf", value);
-    /* Now remove trailing zeroes after the '.' */
-    if (strchr(buf,'.') != NULL) {
-        char *p = buf+len-1;
-        while(*p == '0') {
-            p--;
-            len--;
+    if (isinf(value)) {
+        /* Libc in odd systems (Hi Solaris!) will format infinite in a
+         * different way, so better to handle it in an explicit way. */
+        if (value > 0) {
+            memcpy(buf,"inf",3);
+            len = 3;
+        } else {
+            memcpy(buf,"-inf",4);
+            len = 4;
         }
-        if (*p == '.') len--;
+    } else if (humanfriendly) {
+        /* We use 17 digits precision since with 128 bit floats that precision
+         * after rounding is able to represent most small decimal numbers in a
+         * way that is "non surprising" for the user (that is, most small
+         * decimal numbers will be represented in a way that when converted
+         * back into a string are exactly the same as what the user typed.) */
+        len = snprintf(buf,sizeof(buf),"%.17Lf", value);
+        /* Now remove trailing zeroes after the '.' */
+        if (strchr(buf,'.') != NULL) {
+            char *p = buf+len-1;
+            while(*p == '0') {
+                p--;
+                len--;
+            }
+            if (*p == '.') len--;
+        }
+    } else {
+        len = snprintf(buf,sizeof(buf),"%.17Lg", value);
     }
     return createStringObject(buf,len);
 }
@@ -373,7 +391,9 @@ robj *tryObjectEncoding(robj *o) {
          * Note that we avoid using shared integers when maxmemory is used
          * because every object needs to have a private LRU field for the LRU
          * algorithm to work well. */
-        if (server.maxmemory == 0 &&
+        if ((server.maxmemory == 0 ||
+             (server.maxmemory_policy != REDIS_MAXMEMORY_VOLATILE_LRU &&
+              server.maxmemory_policy != REDIS_MAXMEMORY_ALLKEYS_LRU)) &&
             value >= 0 &&
             value < REDIS_SHARED_INTEGERS)
         {
@@ -696,7 +716,7 @@ robj *objectCommandLookupOrReply(redisClient *c, robj *key, robj *reply) {
 }
 
 /* Object command allows to inspect the internals of an Redis Object.
- * Usage: OBJECT <verb> ... arguments ... */
+ * Usage: OBJECT <refcount|encoding|idletime> <key> */
 void objectCommand(redisClient *c) {
     robj *o;
 

src/pubsub.c

@@ -47,6 +47,12 @@ int listMatchPubsubPattern(void *a, void *b) {
            (equalStringObjects(pa->pattern,pb->pattern));
 }
 
+/* Return the number of channels + patterns a client is subscribed to. */
+int clientSubscriptionsCount(redisClient *c) {
+    return dictSize(c->pubsub_channels)+
+           listLength(c->pubsub_patterns);
+}
+
 /* Subscribe a client to a channel. Returns 1 if the operation succeeded, or
  * 0 if the client was already subscribed to that channel. */
 int pubsubSubscribeChannel(redisClient *c, robj *channel) {
@@ -73,7 +79,7 @@ int pubsubSubscribeChannel(redisClient *c, robj *channel) {
     addReply(c,shared.mbulkhdr[3]);
     addReply(c,shared.subscribebulk);
     addReplyBulk(c,channel);
-    addReplyLongLong(c,dictSize(c->pubsub_channels)+listLength(c->pubsub_patterns));
+    addReplyLongLong(c,clientSubscriptionsCount(c));
     return retval;
 }
 
@@ -135,7 +141,7 @@ int pubsubSubscribePattern(redisClient *c, robj *pattern) {
     addReply(c,shared.mbulkhdr[3]);
     addReply(c,shared.psubscribebulk);
     addReplyBulk(c,pattern);
-    addReplyLongLong(c,dictSize(c->pubsub_channels)+listLength(c->pubsub_patterns));
+    addReplyLongLong(c,clientSubscriptionsCount(c));
     return retval;
 }
 
@@ -168,7 +174,7 @@ int pubsubUnsubscribePattern(redisClient *c, robj *pattern, int notify) {
 }
 
 /* Unsubscribe from all the channels. Return the number of channels the
- * client was subscribed from. */
+ * client was subscribed to. */
 int pubsubUnsubscribeAllChannels(redisClient *c, int notify) {
     dictIterator *di = dictGetSafeIterator(c->pubsub_channels);
     dictEntry *de;
@@ -273,6 +279,7 @@ void subscribeCommand(redisClient *c) {
 
     for (j = 1; j < c->argc; j++)
         pubsubSubscribeChannel(c,c->argv[j]);
+    c->flags |= REDIS_PUBSUB;
 }
 
 void unsubscribeCommand(redisClient *c) {
@@ -284,6 +291,7 @@ void unsubscribeCommand(redisClient *c) {
         for (j = 1; j < c->argc; j++)
             pubsubUnsubscribeChannel(c,c->argv[j],1);
     }
+    if (clientSubscriptionsCount(c) == 0) c->flags &= ~REDIS_PUBSUB;
 }
 
 void psubscribeCommand(redisClient *c) {
@@ -291,6 +299,7 @@ void psubscribeCommand(redisClient *c) {
 
     for (j = 1; j < c->argc; j++)
         pubsubSubscribePattern(c,c->argv[j]);
+    c->flags |= REDIS_PUBSUB;
 }
 
 void punsubscribeCommand(redisClient *c) {
@@ -302,6 +311,7 @@ void punsubscribeCommand(redisClient *c) {
         for (j = 1; j < c->argc; j++)
             pubsubUnsubscribePattern(c,c->argv[j],1);
     }
+    if (clientSubscriptionsCount(c) == 0) c->flags &= ~REDIS_PUBSUB;
 }
 
 void publishCommand(redisClient *c) {
@@ -348,7 +358,7 @@ void pubsubCommand(redisClient *c) {
             list *l = dictFetchValue(server.pubsub_channels,c->argv[j]);
 
             addReplyBulk(c,c->argv[j]);
-            addReplyBulkLongLong(c,l ? listLength(l) : 0);
+            addReplyLongLong(c,l ? listLength(l) : 0);
         }
     } else if (!strcasecmp(c->argv[1]->ptr,"numpat") && c->argc == 2) {
         /* PUBSUB NUMPAT */

src/rand.c

@@ -66,7 +66,7 @@
 #define HI_BIT	(1L << (2 * N - 1))
 
 static uint32_t x[3] = { X0, X1, X2 }, a[3] = { A0, A1, A2 }, c = C;
-static void next();
+static void next(void);
 
 int32_t redisLrand48() {
     next();
@@ -77,7 +77,7 @@ void redisSrand48(int32_t seedval) {
     SEED(X0, LOW(seedval), HIGH(seedval));
 }
 
-static void next() {
+static void next(void) {
     uint32_t p[2], q[2], r[2], carry0, carry1;
 
     MUL(a[0], x[0], p);

src/rdb.c

@@ -333,7 +333,7 @@ int rdbSaveStringObject(rio *rdb, robj *obj) {
 robj *rdbGenericLoadStringObject(rio *rdb, int encode) {
     int isencoded;
     uint32_t len;
-    sds val;
+    robj *o;
 
     len = rdbLoadLen(rdb,&isencoded);
     if (isencoded) {
@@ -350,12 +350,13 @@ robj *rdbGenericLoadStringObject(rio *rdb, int encode) {
     }
 
     if (len == REDIS_RDB_LENERR) return NULL;
-    val = sdsnewlen(NULL,len);
-    if (len && rioRead(rdb,val,len) == 0) {
-        sdsfree(val);
+    o = encode ? createStringObject(NULL,len) :
+                 createRawStringObject(NULL,len);
+    if (len && rioRead(rdb,o->ptr,len) == 0) {
+        decrRefCount(o);
         return NULL;
     }
-    return createObject(REDIS_STRING,val);
+    return o;
 }
 
 robj *rdbLoadStringObject(rio *rdb) {
@@ -474,7 +475,7 @@ int rdbLoadObjectType(rio *rdb) {
     return type;
 }
 
-/* Save a Redis object. Returns -1 on error, 0 on success. */
+/* Save a Redis object. Returns -1 on error, number of bytes written on success. */
 int rdbSaveObject(rio *rdb, robj *o) {
     int n, nwritten = 0;
 
@@ -626,17 +627,100 @@ int rdbSaveKeyValuePair(rio *rdb, robj *key, robj *val,
     return 1;
 }
 
-/* Save the DB on disk. Return REDIS_ERR on error, REDIS_OK on success */
-int rdbSave(char *filename) {
+/* Produces a dump of the database in RDB format sending it to the specified
+ * Redis I/O channel. On success REDIS_OK is returned, otherwise REDIS_ERR
+ * is returned and part of the output, or all the output, can be
+ * missing because of I/O errors.
+ *
+ * When the function returns REDIS_ERR and if 'error' is not NULL, the
+ * integer pointed by 'error' is set to the value of errno just after the I/O
+ * error. */
+int rdbSaveRio(rio *rdb, int *error) {
     dictIterator *di = NULL;
     dictEntry *de;
-    char tmpfile[256];
     char magic[10];
     int j;
     long long now = mstime();
+    uint64_t cksum;
+
+    if (server.rdb_checksum)
+        rdb->update_cksum = rioGenericUpdateChecksum;
+    snprintf(magic,sizeof(magic),"REDIS%04d",REDIS_RDB_VERSION);
+    if (rdbWriteRaw(rdb,magic,9) == -1) goto werr;
+
+    for (j = 0; j < server.dbnum; j++) {
+        redisDb *db = server.db+j;
+        dict *d = db->dict;
+        if (dictSize(d) == 0) continue;
+        di = dictGetSafeIterator(d);
+        if (!di) return REDIS_ERR;
+
+        /* Write the SELECT DB opcode */
+        if (rdbSaveType(rdb,REDIS_RDB_OPCODE_SELECTDB) == -1) goto werr;
+        if (rdbSaveLen(rdb,j) == -1) goto werr;
+
+        /* Iterate this DB writing every entry */
+        while((de = dictNext(di)) != NULL) {
+            sds keystr = dictGetKey(de);
+            robj key, *o = dictGetVal(de);
+            long long expire;
+
+            initStaticStringObject(key,keystr);
+            expire = getExpire(db,&key);
+            if (rdbSaveKeyValuePair(rdb,&key,o,expire,now) == -1) goto werr;
+        }
+        dictReleaseIterator(di);
+    }
+    di = NULL; /* So that we don't release it again on error. */
+
+    /* EOF opcode */
+    if (rdbSaveType(rdb,REDIS_RDB_OPCODE_EOF) == -1) goto werr;
+
+    /* CRC64 checksum. It will be zero if checksum computation is disabled, the
+     * loading code skips the check in this case. */
+    cksum = rdb->cksum;
+    memrev64ifbe(&cksum);
+    if (rioWrite(rdb,&cksum,8) == 0) goto werr;
+    return REDIS_OK;
+
+werr:
+    if (error) *error = errno;
+    if (di) dictReleaseIterator(di);
+    return REDIS_ERR;
+}
+
+/* This is just a wrapper to rdbSaveRio() that additionally adds a prefix
+ * and a suffix to the generated RDB dump. The prefix is:
+ *
+ * $EOF:<40 bytes unguessable hex string>\r\n
+ *
+ * While the suffix is the 40 bytes hex string we announced in the prefix.
+ * This way processes receiving the payload can understand when it ends
+ * without doing any processing of the content. */
+int rdbSaveRioWithEOFMark(rio *rdb, int *error) {
+    char eofmark[REDIS_EOF_MARK_SIZE];
+
+    getRandomHexChars(eofmark,REDIS_EOF_MARK_SIZE);
+    if (error) *error = 0;
+    if (rioWrite(rdb,"$EOF:",5) == 0) goto werr;
+    if (rioWrite(rdb,eofmark,REDIS_EOF_MARK_SIZE) == 0) goto werr;
+    if (rioWrite(rdb,"\r\n",2) == 0) goto werr;
+    if (rdbSaveRio(rdb,error) == REDIS_ERR) goto werr;
+    if (rioWrite(rdb,eofmark,REDIS_EOF_MARK_SIZE) == 0) goto werr;
+    return REDIS_OK;
+
+werr: /* Write error. */
+    /* Set 'error' only if not already set by rdbSaveRio() call. */
+    if (error && *error == 0) *error = errno;
+    return REDIS_ERR;
+}
+
+/* Save the DB on disk. Return REDIS_ERR on error, REDIS_OK on success. */
+int rdbSave(char *filename) {
+    char tmpfile[256];
     FILE *fp;
     rio rdb;
-    uint64_t cksum;
+    int error;
 
     snprintf(tmpfile,256,"temp-%d.rdb", (int) getpid());
     fp = fopen(tmpfile,"w");
@@ -647,47 +731,10 @@ int rdbSave(char *filename) {
     }
 
     rioInitWithFile(&rdb,fp);
-    if (server.rdb_checksum)
-        rdb.update_cksum = rioGenericUpdateChecksum;
-    snprintf(magic,sizeof(magic),"REDIS%04d",REDIS_RDB_VERSION);
-    if (rdbWriteRaw(&rdb,magic,9) == -1) goto werr;
-
-    for (j = 0; j < server.dbnum; j++) {
-        redisDb *db = server.db+j;
-        dict *d = db->dict;
-        if (dictSize(d) == 0) continue;
-        di = dictGetSafeIterator(d);
-        if (!di) {
-            fclose(fp);
-            return REDIS_ERR;
-        }
-
-        /* Write the SELECT DB opcode */
-        if (rdbSaveType(&rdb,REDIS_RDB_OPCODE_SELECTDB) == -1) goto werr;
-        if (rdbSaveLen(&rdb,j) == -1) goto werr;
-
-        /* Iterate this DB writing every entry */
-        while((de = dictNext(di)) != NULL) {
-            sds keystr = dictGetKey(de);
-            robj key, *o = dictGetVal(de);
-            long long expire;
-
-            initStaticStringObject(key,keystr);
-            expire = getExpire(db,&key);
-            if (rdbSaveKeyValuePair(&rdb,&key,o,expire,now) == -1) goto werr;
-        }
-        dictReleaseIterator(di);
+    if (rdbSaveRio(&rdb,&error) == REDIS_ERR) {
+        errno = error;
+        goto werr;
     }
-    di = NULL; /* So that we don't release it again on error. */
-
-    /* EOF opcode */
-    if (rdbSaveType(&rdb,REDIS_RDB_OPCODE_EOF) == -1) goto werr;
-
-    /* CRC64 checksum. It will be zero if checksum computation is disabled, the
-     * loading code skips the check in this case. */
-    cksum = rdb.cksum;
-    memrev64ifbe(&cksum);
-    rioWrite(&rdb,&cksum,8);
 
     /* Make sure data will not remain on the OS's output buffers */
     if (fflush(fp) == EOF) goto werr;
@@ -711,7 +758,6 @@ werr:
     fclose(fp);
     unlink(tmpfile);
     redisLog(REDIS_WARNING,"Write error saving DB on disk: %s", strerror(errno));
-    if (di) dictReleaseIterator(di);
     return REDIS_ERR;
 }
 
@@ -756,6 +802,7 @@ int rdbSaveBackground(char *filename) {
         redisLog(REDIS_NOTICE,"Background saving started by pid %d",childpid);
         server.rdb_save_time_start = time(NULL);
         server.rdb_child_pid = childpid;
+        server.rdb_child_type = REDIS_RDB_CHILD_TYPE_DISK;
         updateDictResizePolicy();
         return REDIS_OK;
     }
@@ -765,7 +812,7 @@ int rdbSaveBackground(char *filename) {
 void rdbRemoveTempFile(pid_t childpid) {
     char tmpfile[256];
 
-    snprintf(tmpfile,256,"temp-%d.rdb", (int) childpid);
+    snprintf(tmpfile,sizeof(tmpfile),"temp-%d.rdb", (int) childpid);
     unlink(tmpfile);
 }
 
@@ -942,7 +989,7 @@ robj *rdbLoadObject(int rdbtype, rio *rdb) {
 
             /* Add pair to hash table */
             ret = dictAdd((dict*)o->ptr, field, value);
-            redisAssert(ret == REDIS_OK);
+            redisAssert(ret == DICT_OK);
         }
 
         /* All pairs should be read by now */
@@ -1190,8 +1237,9 @@ eoferr: /* unexpected end of file is handled here with a fatal exit */
     return REDIS_ERR; /* Just to avoid warning */
 }
 
-/* A background saving child (BGSAVE) terminated its work. Handle this. */
-void backgroundSaveDoneHandler(int exitcode, int bysignal) {
+/* A background saving child (BGSAVE) terminated its work. Handle this.
+ * This function covers the case of actual BGSAVEs. */
+void backgroundSaveDoneHandlerDisk(int exitcode, int bysignal) {
     if (!bysignal && exitcode == 0) {
         redisLog(REDIS_NOTICE,
             "Background saving terminated with success");
@@ -1216,11 +1264,258 @@ void backgroundSaveDoneHandler(int exitcode, int bysignal) {
             server.lastbgsave_status = REDIS_ERR;
     }
     server.rdb_child_pid = -1;
+    server.rdb_child_type = REDIS_RDB_CHILD_TYPE_NONE;
     server.rdb_save_time_last = time(NULL)-server.rdb_save_time_start;
     server.rdb_save_time_start = -1;
     /* Possibly there are slaves waiting for a BGSAVE in order to be served
      * (the first stage of SYNC is a bulk transfer of dump.rdb) */
-    updateSlavesWaitingBgsave(exitcode == 0 ? REDIS_OK : REDIS_ERR);
+    updateSlavesWaitingBgsave((!bysignal && exitcode == 0) ? REDIS_OK : REDIS_ERR, REDIS_RDB_CHILD_TYPE_DISK);
+}
+
+/* A background saving child (BGSAVE) terminated its work. Handle this.
+ * This function covers the case of RDB -> Salves socket transfers for
+ * diskless replication. */
+void backgroundSaveDoneHandlerSocket(int exitcode, int bysignal) {
+    uint64_t *ok_slaves;
+
+    if (!bysignal && exitcode == 0) {
+        redisLog(REDIS_NOTICE,
+            "Background RDB transfer terminated with success");
+    } else if (!bysignal && exitcode != 0) {
+        redisLog(REDIS_WARNING, "Background transfer error");
+    } else {
+        redisLog(REDIS_WARNING,
+            "Background transfer terminated by signal %d", bysignal);
+    }
+    server.rdb_child_pid = -1;
+    server.rdb_child_type = REDIS_RDB_CHILD_TYPE_NONE;
+    server.rdb_save_time_start = -1;
+
+    /* If the child returns an OK exit code, read the set of slave client
+     * IDs and the associated status code. We'll terminate all the slaves
+     * in error state.
+     *
+     * If the process returned an error, consider the list of slaves that
+     * can continue to be emtpy, so that it's just a special case of the
+     * normal code path. */
+    ok_slaves = zmalloc(sizeof(uint64_t)); /* Make space for the count. */
+    ok_slaves[0] = 0;
+    if (!bysignal && exitcode == 0) {
+        int readlen = sizeof(uint64_t);
+
+        if (read(server.rdb_pipe_read_result_from_child, ok_slaves, readlen) ==
+                 readlen)
+        {
+            readlen = ok_slaves[0]*sizeof(uint64_t)*2;
+
+            /* Make space for enough elements as specified by the first
+             * uint64_t element in the array. */
+            ok_slaves = zrealloc(ok_slaves,sizeof(uint64_t)+readlen);
+            if (readlen &&
+                read(server.rdb_pipe_read_result_from_child, ok_slaves+1,
+                     readlen) != readlen)
+            {
+                ok_slaves[0] = 0;
+            }
+        }
+    }
+
+    close(server.rdb_pipe_read_result_from_child);
+    close(server.rdb_pipe_write_result_to_parent);
+
+    /* We can continue the replication process with all the slaves that
+     * correctly received the full payload. Others are terminated. */
+    listNode *ln;
+    listIter li;
+
+    listRewind(server.slaves,&li);
+    while((ln = listNext(&li))) {
+        redisClient *slave = ln->value;
+
+        if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_END) {
+            uint64_t j;
+            int errorcode = 0;
+
+            /* Search for the slave ID in the reply. In order for a slave to
+             * continue the replication process, we need to find it in the list,
+             * and it must have an error code set to 0 (which means success). */
+            for (j = 0; j < ok_slaves[0]; j++) {
+                if (slave->id == ok_slaves[2*j+1]) {
+                    errorcode = ok_slaves[2*j+2];
+                    break; /* Found in slaves list. */
+                }
+            }
+            if (j == ok_slaves[0] || errorcode != 0) {
+                redisLog(REDIS_WARNING,
+                "Closing slave %s: child->slave RDB transfer failed: %s",
+                    replicationGetSlaveName(slave),
+                    (errorcode == 0) ? "RDB transfer child aborted"
+                                     : strerror(errorcode));
+                freeClient(slave);
+            } else {
+                redisLog(REDIS_WARNING,
+                "Slave %s correctly received the streamed RDB file.",
+                    replicationGetSlaveName(slave));
+                /* Restore the socket as non-blocking. */
+                anetNonBlock(NULL,slave->fd);
+                anetSendTimeout(NULL,slave->fd,0);
+            }
+        }
+    }
+    zfree(ok_slaves);
+
+    updateSlavesWaitingBgsave((!bysignal && exitcode == 0) ? REDIS_OK : REDIS_ERR, REDIS_RDB_CHILD_TYPE_SOCKET);
+}
+
+/* When a background RDB saving/transfer terminates, call the right handler. */
+void backgroundSaveDoneHandler(int exitcode, int bysignal) {
+    switch(server.rdb_child_type) {
+    case REDIS_RDB_CHILD_TYPE_DISK:
+        backgroundSaveDoneHandlerDisk(exitcode,bysignal);
+        break;
+    case REDIS_RDB_CHILD_TYPE_SOCKET:
+        backgroundSaveDoneHandlerSocket(exitcode,bysignal);
+        break;
+    default:
+        redisPanic("Unknown RDB child type.");
+        break;
+    }
+}
+
+/* Spawn an RDB child that writes the RDB to the sockets of the slaves
+ * that are currently in REDIS_REPL_WAIT_BGSAVE_START state. */
+int rdbSaveToSlavesSockets(void) {
+    int *fds;
+    uint64_t *clientids;
+    int numfds;
+    listNode *ln;
+    listIter li;
+    pid_t childpid;
+    long long start;
+    int pipefds[2];
+
+    if (server.rdb_child_pid != -1) return REDIS_ERR;
+
+    /* Before to fork, create a pipe that will be used in order to
+     * send back to the parent the IDs of the slaves that successfully
+     * received all the writes. */
+    if (pipe(pipefds) == -1) return REDIS_ERR;
+    server.rdb_pipe_read_result_from_child = pipefds[0];
+    server.rdb_pipe_write_result_to_parent = pipefds[1];
+
+    /* Collect the file descriptors of the slaves we want to transfer
+     * the RDB to, which are i WAIT_BGSAVE_START state. */
+    fds = zmalloc(sizeof(int)*listLength(server.slaves));
+    /* We also allocate an array of corresponding client IDs. This will
+     * be useful for the child process in order to build the report
+     * (sent via unix pipe) that will be sent to the parent. */
+    clientids = zmalloc(sizeof(uint64_t)*listLength(server.slaves));
+    numfds = 0;
+
+    listRewind(server.slaves,&li);
+    while((ln = listNext(&li))) {
+        redisClient *slave = ln->value;
+
+        if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_START) {
+            clientids[numfds] = slave->id;
+            fds[numfds++] = slave->fd;
+            slave->replstate = REDIS_REPL_WAIT_BGSAVE_END;
+            /* Put the socket in non-blocking mode to simplify RDB transfer.
+             * We'll restore it when the children returns (since duped socket
+             * will share the O_NONBLOCK attribute with the parent). */
+            anetBlock(NULL,slave->fd);
+            anetSendTimeout(NULL,slave->fd,server.repl_timeout*1000);
+        }
+    }
+
+    /* Create the child process. */
+    start = ustime();
+    if ((childpid = fork()) == 0) {
+        /* Child */
+        int retval;
+        rio slave_sockets;
+
+        rioInitWithFdset(&slave_sockets,fds,numfds);
+        zfree(fds);
+
+        closeListeningSockets(0);
+        redisSetProcTitle("redis-rdb-to-slaves");
+
+        retval = rdbSaveRioWithEOFMark(&slave_sockets,NULL);
+        if (retval == REDIS_OK && rioFlush(&slave_sockets) == 0)
+            retval = REDIS_ERR;
+
+        if (retval == REDIS_OK) {
+            size_t private_dirty = zmalloc_get_private_dirty();
+
+            if (private_dirty) {
+                redisLog(REDIS_NOTICE,
+                    "RDB: %zu MB of memory used by copy-on-write",
+                    private_dirty/(1024*1024));
+            }
+
+            /* If we are returning OK, at least one slave was served
+             * with the RDB file as expected, so we need to send a report
+             * to the parent via the pipe. The format of the message is:
+             *
+             * <len> <slave[0].id> <slave[0].error> ...
+             *
+             * len, slave IDs, and slave errors, are all uint64_t integers,
+             * so basically the reply is composed of 64 bits for the len field
+             * plus 2 additional 64 bit integers for each entry, for a total
+             * of 'len' entries.
+             *
+             * The 'id' represents the slave's client ID, so that the master
+             * can match the report with a specific slave, and 'error' is
+             * set to 0 if the replication process terminated with a success
+             * or the error code if an error occurred. */
+            void *msg = zmalloc(sizeof(uint64_t)*(1+2*numfds));
+            uint64_t *len = msg;
+            uint64_t *ids = len+1;
+            int j, msglen;
+
+            *len = numfds;
+            for (j = 0; j < numfds; j++) {
+                *ids++ = clientids[j];
+                *ids++ = slave_sockets.io.fdset.state[j];
+            }
+
+            /* Write the message to the parent. If we have no good slaves or
+             * we are unable to transfer the message to the parent, we exit
+             * with an error so that the parent will abort the replication
+             * process with all the childre that were waiting. */
+            msglen = sizeof(uint64_t)*(1+2*numfds);
+            if (*len == 0 ||
+                write(server.rdb_pipe_write_result_to_parent,msg,msglen)
+                != msglen)
+            {
+                retval = REDIS_ERR;
+            }
+        }
+        exitFromChild((retval == REDIS_OK) ? 0 : 1);
+    } else {
+        /* Parent */
+        zfree(clientids); /* Not used by parent. Free ASAP. */
+        server.stat_fork_time = ustime()-start;
+        server.stat_fork_rate = (double) zmalloc_used_memory() * 1000000 / server.stat_fork_time / (1024*1024*1024); /* GB per second. */
+        latencyAddSampleIfNeeded("fork",server.stat_fork_time/1000);
+        if (childpid == -1) {
+            redisLog(REDIS_WARNING,"Can't save in background: fork: %s",
+                strerror(errno));
+            zfree(fds);
+            close(pipefds[0]);
+            close(pipefds[1]);
+            return REDIS_ERR;
+        }
+        redisLog(REDIS_NOTICE,"Background RDB transfer started by pid %d",childpid);
+        server.rdb_save_time_start = time(NULL);
+        server.rdb_child_pid = childpid;
+        server.rdb_child_type = REDIS_RDB_CHILD_TYPE_SOCKET;
+        updateDictResizePolicy();
+        zfree(fds);
+        return REDIS_OK;
+    }
+    return REDIS_OK; /* unreached */
 }
 
 void saveCommand(redisClient *c) {

src/rdb.h

@@ -101,6 +101,7 @@ int rdbSaveObjectType(rio *rdb, robj *o);
 int rdbLoadObjectType(rio *rdb);
 int rdbLoad(char *filename);
 int rdbSaveBackground(char *filename);
+int rdbSaveToSlavesSockets(void);
 void rdbRemoveTempFile(pid_t childpid);
 int rdbSave(char *filename);
 int rdbSaveObject(rio *rdb, robj *o);

src/redis-benchmark.c

@@ -77,6 +77,7 @@ static struct config {
     int dbnum;
     sds dbnumstr;
     char *tests;
+    char *auth;
 } config;
 
 typedef struct _client {
@@ -213,7 +214,7 @@ static void readHandler(aeEventLoop *el, int fd, void *privdata, int mask) {
                 freeReplyObject(reply);
 
                 if (c->selectlen) {
-                    int j;
+                    size_t j;
 
                     /* This is the OK from SELECT. Just discard the SELECT
                      * from the buffer. */
@@ -325,6 +326,13 @@ static client createClient(char *cmd, size_t len, client from) {
      * the example client buffer. */
     c->obuf = sdsempty();
 
+    if (config.auth) {
+        char *buf = NULL;
+        int len = redisFormatCommand(&buf, "AUTH %s", config.auth);
+        c->obuf = sdscatlen(c->obuf, buf, len);
+        free(buf);
+    }
+
     /* If a DB number different than zero is selected, prefix our request
      * buffer with the SELECT command, that will be discarded the first
      * time the replies are received, so if the client is reused the
@@ -346,6 +354,7 @@ static client createClient(char *cmd, size_t len, client from) {
         for (j = 0; j < config.pipeline; j++)
             c->obuf = sdscatlen(c->obuf,cmd,len);
     }
+
     c->written = 0;
     c->pending = config.pipeline;
     c->randptr = NULL;
@@ -359,7 +368,7 @@ static client createClient(char *cmd, size_t len, client from) {
             c->randfree = 0;
             c->randptr = zmalloc(sizeof(char*)*c->randlen);
             /* copy the offsets. */
-            for (j = 0; j < c->randlen; j++) {
+            for (j = 0; j < (int)c->randlen; j++) {
                 c->randptr[j] = c->obuf + (from->randptr[j]-from->obuf);
                 /* Adjust for the different select prefix length. */
                 c->randptr[j] += c->selectlen - from->selectlen;
@@ -389,15 +398,6 @@ static client createClient(char *cmd, size_t len, client from) {
 
 static void createMissingClients(client c) {
     int n = 0;
-    char *buf = c->obuf;
-    size_t buflen = sdslen(c->obuf);
-
-    /* If we are cloning from a client with a SELECT prefix, skip it since the
-     * client will be created with the prefixed SELECT if needed. */
-    if (c->selectlen) {
-        buf += c->selectlen;
-        buflen -= c->selectlen;
-    }
 
     while(config.liveclients < config.numclients) {
         createClient(NULL,0,c);
@@ -489,6 +489,9 @@ int parseOptions(int argc, const char **argv) {
         } else if (!strcmp(argv[i],"-s")) {
             if (lastarg) goto invalid;
             config.hostsocket = strdup(argv[++i]);
+        } else if (!strcmp(argv[i],"-a") ) {
+            if (lastarg) goto invalid;
+            config.auth = strdup(argv[++i]);
         } else if (!strcmp(argv[i],"-d")) {
             if (lastarg) goto invalid;
             config.datasize = atoi(argv[++i]);
@@ -550,8 +553,9 @@ usage:
 " -h <hostname>      Server hostname (default 127.0.0.1)\n"
 " -p <port>          Server port (default 6379)\n"
 " -s <socket>        Server socket (overrides host and port)\n"
+" -a <password>      Password for Redis Auth\n"
 " -c <clients>       Number of parallel connections (default 50)\n"
-" -n <requests>      Total number of requests (default 10000)\n"
+" -n <requests>      Total number of requests (default 100000)\n"
 " -d <size>          Data size of SET/GET value in bytes (default 2)\n"
 " -dbnum <db>        SELECT the specified db number (default 0)\n"
 " -k <boolean>       1=keep alive 0=reconnect (default 1)\n"
@@ -593,7 +597,7 @@ int showThroughput(struct aeEventLoop *eventLoop, long long id, void *clientData
     REDIS_NOTUSED(clientData);
 
     if (config.liveclients == 0) {
-        fprintf(stderr,"All clients disconnected... aborting.");
+        fprintf(stderr,"All clients disconnected... aborting.\n");
         exit(1);
     }
 
@@ -631,7 +635,7 @@ int main(int argc, const char **argv) {
     signal(SIGPIPE, SIG_IGN);
 
     config.numclients = 50;
-    config.requests = 10000;
+    config.requests = 100000;
     config.liveclients = 0;
     config.el = aeCreateEventLoop(1024*10);
     aeCreateTimeEvent(config.el,1,showThroughput,NULL,NULL);
@@ -651,6 +655,7 @@ int main(int argc, const char **argv) {
     config.hostsocket = NULL;
     config.tests = NULL;
     config.dbnum = 0;
+    config.auth = NULL;
 
     i = parseOptions(argc,argv);
     argc -= i;
@@ -688,8 +693,8 @@ int main(int argc, const char **argv) {
     }
 
     /* Run default benchmark suite. */
+    data = zmalloc(config.datasize+1);
     do {
-        data = zmalloc(config.datasize+1);
         memset(data,'x',config.datasize);
         data[config.datasize] = '\0';
 

src/redis-check-dump.c

@@ -138,8 +138,10 @@ typedef struct {
  * at runtime to avoid strange compiler optimizations. */
 static double R_Zero, R_PosInf, R_NegInf, R_Nan;
 
+#define MAX_TYPES_NUM 256
+#define MAX_TYPE_NAME_LEN 16
 /* store string types for output */
-static char types[256][16];
+static char types[MAX_TYPES_NUM][MAX_TYPE_NAME_LEN];
 
 /* Return true if 't' is a valid object type. */
 int checkType(unsigned char t) {
@@ -166,7 +168,7 @@ int readBytes(void *target, long num) {
     return 1;
 }
 
-int processHeader() {
+int processHeader(void) {
     char buf[10] = "_________";
     int dump_version;
 
@@ -335,6 +337,7 @@ char* loadStringObject() {
     if (len == REDIS_RDB_LENERR) return NULL;
 
     char *buf = malloc(sizeof(char) * (len+1));
+    if (buf == NULL) return NULL;
     buf[len] = '\0';
     if (!readBytes(buf, len)) {
         free(buf);
@@ -600,7 +603,7 @@ void printErrorStack(entry *e) {
     }
 }
 
-void process() {
+void process(void) {
     uint64_t num_errors = 0, num_valid_ops = 0, num_valid_bytes = 0;
     entry entry;
     int dump_version = processHeader();
@@ -611,7 +614,7 @@ void process() {
             printf("RDB version >= 5 but no room for checksum.\n");
             exit(1);
         }
-        positions[0].size -= 8;;
+        positions[0].size -= 8;
     }
 
     level = 1;

src/redis-cli.c

@@ -94,10 +94,11 @@ static struct config {
     sds mb_delim;
     char prompt[128];
     char *eval;
+    int last_cmd_type;
 } config;
 
 static volatile sig_atomic_t force_cancel_loop = 0;
-static void usage();
+static void usage(void);
 static void slaveMode(void);
 char *redisGitSHA1(void);
 char *redisGitDirty(void);
@@ -131,7 +132,7 @@ static void cliRefreshPrompt(void) {
                        strchr(config.hostip,':') ? "[%s]:%d" : "%s:%d",
                        config.hostip, config.hostport);
     /* Add [dbnum] if needed */
-    if (config.dbnum != 0)
+    if (config.dbnum != 0 && config.last_cmd_type != REDIS_REPLY_ERROR)
         len += snprintf(config.prompt+len,sizeof(config.prompt)-len,"[%d]",
             config.dbnum);
     snprintf(config.prompt+len,sizeof(config.prompt)-len,"> ");
@@ -157,7 +158,7 @@ typedef struct {
 static helpEntry *helpEntries;
 static int helpEntriesLen;
 
-static sds cliVersion() {
+static sds cliVersion(void) {
     sds version;
     version = sdscatprintf(sdsempty(), "%s", REDIS_VERSION);
 
@@ -171,7 +172,7 @@ static sds cliVersion() {
     return version;
 }
 
-static void cliInitHelp() {
+static void cliInitHelp(void) {
     int commandslen = sizeof(commandHelp)/sizeof(struct commandHelp);
     int groupslen = sizeof(commandGroups)/sizeof(char*);
     int i, len, pos = 0;
@@ -210,7 +211,7 @@ static void cliOutputCommandHelp(struct commandHelp *help, int group) {
 }
 
 /* Print generic help. */
-static void cliOutputGenericHelp() {
+static void cliOutputGenericHelp(void) {
     sds version = cliVersion();
     printf(
         "redis-cli %s\r\n"
@@ -320,8 +321,10 @@ static int cliSelect() {
 
     reply = redisCommand(context,"SELECT %d",config.dbnum);
     if (reply != NULL) {
+        int result = REDIS_OK;
+        if (reply->type == REDIS_REPLY_ERROR) result = REDIS_ERR;
         freeReplyObject(reply);
-        return REDIS_OK;
+        return result;
     }
     return REDIS_ERR;
 }
@@ -365,7 +368,7 @@ static int cliConnect(int force) {
     return REDIS_OK;
 }
 
-static void cliPrintContextError() {
+static void cliPrintContextError(void) {
     if (context == NULL) return;
     fprintf(stderr,"Error: %s\n",context->errstr);
 }
@@ -514,11 +517,15 @@ static int cliReadReply(int output_raw_strings) {
     int output = 1;
 
     if (redisGetReply(context,&_reply) != REDIS_OK) {
-        if (config.shutdown)
+        if (config.shutdown) {
+            redisFree(context);
+            context = NULL;
             return REDIS_OK;
+        }
         if (config.interactive) {
             /* Filter cases where we should reconnect */
-            if (context->err == REDIS_ERR_IO && errno == ECONNRESET)
+            if (context->err == REDIS_ERR_IO &&
+                (errno == ECONNRESET || errno == EPIPE))
                 return REDIS_ERR;
             if (context->err == REDIS_ERR_EOF)
                 return REDIS_ERR;
@@ -530,6 +537,8 @@ static int cliReadReply(int output_raw_strings) {
 
     reply = (redisReply*)_reply;
 
+    config.last_cmd_type = reply->type;
+
     /* Check if we need to connect to a different node and reissue the
      * request. */
     if (config.cluster_mode && reply->type == REDIS_REPLY_ERROR &&
@@ -639,6 +648,7 @@ static int cliSendCommand(int argc, char **argv, int repeat) {
             printf("Entering slave output mode...  (press Ctrl-C to quit)\n");
             slaveMode();
             config.slave_mode = 0;
+            free(argvlen);
             return REDIS_ERR;  /* Error = slaveMode lost connection to master */
         }
 
@@ -650,6 +660,8 @@ static int cliSendCommand(int argc, char **argv, int repeat) {
             if (!strcasecmp(command,"select") && argc == 2) {
                 config.dbnum = atoi(argv[1]);
                 cliRefreshPrompt();
+            } else if (!strcasecmp(command,"auth") && argc == 2) {
+                cliSelect();
             }
         }
         if (config.interval) usleep(config.interval);
@@ -724,6 +736,8 @@ static int parseOptions(int argc, char **argv) {
             config.auth = argv[++i];
         } else if (!strcmp(argv[i],"--raw")) {
             config.output = OUTPUT_RAW;
+        } else if (!strcmp(argv[i],"--no-raw")) {
+            config.output = OUTPUT_STANDARD;
         } else if (!strcmp(argv[i],"--csv")) {
             config.output = OUTPUT_CSV;
         } else if (!strcmp(argv[i],"--latency")) {
@@ -795,7 +809,7 @@ static sds readArgFromStdin(void) {
     return arg;
 }
 
-static void usage() {
+static void usage(void) {
     sds version = cliVersion();
     fprintf(stderr,
 "redis-cli %s\n"
@@ -814,7 +828,9 @@ static void usage() {
 "  -c                 Enable cluster mode (follow -ASK and -MOVED redirections).\n"
 "  --raw              Use raw formatting for replies (default when STDOUT is\n"
 "                     not a tty).\n"
+"  --no-raw           Force formatted output even when STDOUT is not a tty.\n"
 "  --csv              Output in CSV format.\n"
+"  --stat             Print rolling stats about server: mem, clients, ...\n"
 "  --latency          Enter a special mode continuously sampling latency.\n"
 "  --latency-history  Like --latency but tracking latency changes over time.\n"
 "                     Default time interval is 15 sec. Change it using -i.\n"
@@ -862,8 +878,7 @@ static char **convertToSds(int count, char** args) {
   return sds;
 }
 
-#define LINE_BUFLEN 4096
-static void repl() {
+static void repl(void) {
     sds historyfile = NULL;
     int history = 0;
     char *line;
@@ -1406,7 +1421,7 @@ static int toIntType(char *key, char *type) {
 
 static void getKeyTypes(redisReply *keys, int *types) {
     redisReply *reply;
-    int i;
+    unsigned int i;
 
     /* Pipeline TYPE commands */
     for(i=0;i<keys->elements;i++) {
@@ -1435,7 +1450,7 @@ static void getKeySizes(redisReply *keys, int *types,
 {
     redisReply *reply;
     char *sizecmds[] = {"STRLEN","LLEN","SCARD","HLEN","ZCARD"};
-    int i;
+    unsigned int i;
 
     /* Pipeline size commands */
     for(i=0;i<keys->elements;i++) {
@@ -1482,7 +1497,8 @@ static void findBigKeys(void) {
     char *typename[] = {"string","list","set","hash","zset"};
     char *typeunit[] = {"bytes","items","members","fields","members"};
     redisReply *reply, *keys;
-    int type, *types=NULL, arrsize=0, i;
+    unsigned int arrsize=0, i;
+    int type, *types=NULL;
     double pct;
 
     /* Total keys pre scanning */
@@ -1497,7 +1513,7 @@ static void findBigKeys(void) {
     for(i=0;i<TYPE_NONE; i++) {
         maxkeys[i] = sdsempty();
         if(!maxkeys[i]) {
-            fprintf(stderr, "Failed to allocate memory for largest key names!");
+            fprintf(stderr, "Failed to allocate memory for largest key names!\n");
             exit(1);
         }
     }
@@ -1666,7 +1682,7 @@ void bytesToHuman(char *s, long long n) {
     }
 }
 
-static void statMode() {
+static void statMode(void) {
     redisReply *reply;
     long aux, requests = 0;
     int i = 0;
@@ -1752,7 +1768,7 @@ static void statMode() {
  * Scan mode
  *--------------------------------------------------------------------------- */
 
-static void scanMode() {
+static void scanMode(void) {
     redisReply *reply;
     unsigned long long cur = 0;
 
@@ -1769,7 +1785,7 @@ static void scanMode() {
             printf("ERROR: %s\n", reply->str);
             exit(1);
         } else {
-            int j;
+            unsigned int j;
 
             cur = strtoull(reply->element[0]->str,NULL,10);
             for (j = 0; j < reply->element[1]->elements; j++)
@@ -1840,11 +1856,15 @@ static void intrinsicLatencyMode(void) {
             printf("Max latency so far: %lld microseconds.\n", max_latency);
         }
 
+        double avg_us = (double)run_time/runs;
+        double avg_ns = avg_us * 10e3;
         if (force_cancel_loop || end > test_end) {
-            printf("\n%lld total runs (avg %lld microseconds per run).\n",
-                runs, run_time/runs);
-            printf("Worst run took %.02fx times the average.\n",
-                (double) max_latency / (run_time/runs));
+            printf("\n%lld total runs "
+                "(avg latency: "
+                "%.4f microseconds / %.2f nanoseconds per run).\n",
+                runs, avg_us, avg_ns);
+            printf("Worst run took %.0fx longer than the average latency.\n",
+                max_latency / avg_us);
             exit(0);
         }
     }
@@ -1883,6 +1903,8 @@ int main(int argc, char **argv) {
     config.stdinarg = 0;
     config.auth = NULL;
     config.eval = NULL;
+    config.last_cmd_type = -1;
+
     if (!isatty(fileno(stdout)) && (getenv("FAKETTY") == NULL))
         config.output = OUTPUT_RAW;
     else
@@ -1894,6 +1916,8 @@ int main(int argc, char **argv) {
     argc -= firstarg;
     argv += firstarg;
 
+    signal(SIGPIPE, SIG_IGN);
+
     /* Latency mode */
     if (config.latency_mode) {
         if (cliConnect(0) == REDIS_ERR) exit(1);

src/redis-trib.rb

@@ -50,14 +50,16 @@ end
 class ClusterNode
     def initialize(addr)
         s = addr.split(":")
-        if s.length != 2
-            puts "Invalid node name #{addr}"
-            exit 1
+        if s.length < 2
+           puts "Invalid IP or Port (given as #{addr}) - use IP:Port format"
+           exit 1
         end
+        port = s.pop # removes port from split array
+        ip = s.join(":") # if s.length > 1 here, it's IPv6, so restore address
         @r = nil
         @info = {}
-        @info[:host] = s[0]
-        @info[:port] = s[1]
+        @info[:host] = ip
+        @info[:port] = port
         @info[:slots] = {}
         @info[:migrating] = {}
         @info[:importing] = {}
@@ -686,8 +688,13 @@ class RedisTrib
                     f[:flags].index("fail")
             fnode = ClusterNode.new(f[:addr])
             fnode.connect()
-            fnode.load_info()
-            add_node(fnode)
+            next if !fnode.r
+            begin
+                fnode.load_info()
+                add_node(fnode)
+            rescue => e
+                xputs "[ERR] Unable to load info for node #{fnode}"
+            end
         }
         populate_nodes_replicas_info
     end
@@ -821,50 +828,96 @@ class RedisTrib
             puts "*** Please fix your cluster problems before resharding"
             exit 1
         end
-        numslots = 0
-        while numslots <= 0 or numslots > ClusterHashSlots
-            print "How many slots do you want to move (from 1 to #{ClusterHashSlots})? "
-            numslots = STDIN.gets.to_i
+
+        # Get number of slots
+        if opt['slots']
+            numslots = opt['slots'].to_i
+        else
+            numslots = 0
+            while numslots <= 0 or numslots > ClusterHashSlots
+                print "How many slots do you want to move (from 1 to #{ClusterHashSlots})? "
+                numslots = STDIN.gets.to_i
+            end
         end
-        target = nil
-        while not target
-            print "What is the receiving node ID? "
-            target = get_node_by_name(STDIN.gets.chop)
+
+        # Get the target instance
+        if opt['to']
+            target = get_node_by_name(opt['to'])
             if !target || target.has_flag?("slave")
                 xputs "*** The specified node is not known or not a master, please retry."
-                target = nil
+                exit 1
+            end
+        else
+            target = nil
+            while not target
+                print "What is the receiving node ID? "
+                target = get_node_by_name(STDIN.gets.chop)
+                if !target || target.has_flag?("slave")
+                    xputs "*** The specified node is not known or not a master, please retry."
+                    target = nil
+                end
             end
         end
+
+        # Get the source instances
         sources = []
-        xputs "Please enter all the source node IDs."
-        xputs "  Type 'all' to use all the nodes as source nodes for the hash slots."
-        xputs "  Type 'done' once you entered all the source nodes IDs."
-        while true
-            print "Source node ##{sources.length+1}:"
-            line = STDIN.gets.chop
-            src = get_node_by_name(line)
-            if line == "done"
-                if sources.length == 0
-                    puts "No source nodes given, operation aborted"
-                    exit 1
-                else
+        if opt['from']
+            opt['from'].split(',').each{|node_id|
+                if node_id == "all"
+                    sources = "all"
                     break
                 end
-            elsif line == "all"
-                @nodes.each{|n|
-                    next if n.info[:name] == target.info[:name]
-                    next if n.has_flag?("slave")
-                    sources << n
-                }
-                break
-            elsif !src || src.has_flag?("slave")
-                xputs "*** The specified node is not known or is not a master, please retry."
-            elsif src.info[:name] == target.info[:name]
-                xputs "*** It is not possible to use the target node as source node."
-            else
+                src = get_node_by_name(node_id)
+                if !src || src.has_flag?("slave")
+                    xputs "*** The specified node is not known or is not a master, please retry."
+                    exit 1
+                end
                 sources << src
+            }
+        else
+            xputs "Please enter all the source node IDs."
+            xputs "  Type 'all' to use all the nodes as source nodes for the hash slots."
+            xputs "  Type 'done' once you entered all the source nodes IDs."
+            while true
+                print "Source node ##{sources.length+1}:"
+                line = STDIN.gets.chop
+                src = get_node_by_name(line)
+                if line == "done"
+                    break
+                elsif line == "all"
+                    sources = "all"
+                    break
+                elsif !src || src.has_flag?("slave")
+                    xputs "*** The specified node is not known or is not a master, please retry."
+                elsif src.info[:name] == target.info[:name]
+                    xputs "*** It is not possible to use the target node as source node."
+                else
+                    sources << src
+                end
             end
         end
+
+        if sources.length == 0
+            puts "*** No source nodes given, operation aborted"
+            exit 1
+        end
+
+        # Handle soures == all.
+        if sources == "all"
+            sources = []
+            @nodes.each{|n|
+                next if n.info[:name] == target.info[:name]
+                next if n.has_flag?("slave")
+                sources << n
+            }
+        end
+
+        # Check if the destination node is the same of any source nodes.
+        if sources.index(target)
+            xputs "*** Target node is also listed among the source nodes!"
+            exit 1
+        end
+
         puts "\nReady to move #{numslots} slots."
         puts "  Source nodes:"
         sources.each{|s| puts "    "+s.info_string}
@@ -873,9 +926,11 @@ class RedisTrib
         reshard_table = compute_reshard_table(sources,numslots)
         puts "  Resharding plan:"
         show_reshard_table(reshard_table)
-        print "Do you want to proceed with the proposed reshard plan (yes/no)? "
-        yesno = STDIN.gets.chop
-        exit(1) if (yesno != "yes")
+        if !opt['yes']
+            print "Do you want to proceed with the proposed reshard plan (yes/no)? "
+            yesno = STDIN.gets.chop
+            exit(1) if (yesno != "yes")
+        end
         reshard_table.each{|e|
             move_slot(e[:source],target,e[:slot],:verbose=>true)
         }
@@ -1253,7 +1308,8 @@ COMMANDS={
 ALLOWED_OPTIONS={
     "create" => {"replicas" => true},
     "add-node" => {"slave" => false, "master-id" => true},
-    "import" => {"from" => :required}
+    "import" => {"from" => :required},
+    "reshard" => {"from" => true, "to" => true, "slots" => true, "yes" => false}
 }
 
 def show_help

src/redis.c

@@ -31,6 +31,7 @@
 #include "cluster.h"
 #include "slowlog.h"
 #include "bio.h"
+#include "latency.h"
 
 #include <time.h>
 #include <signal.h>
@@ -66,7 +67,6 @@ double R_Zero, R_PosInf, R_NegInf, R_Nan;
 
 /* Global vars */
 struct redisServer server; /* server global state */
-struct redisCommand *commandTable;
 
 /* Our command table.
  *
@@ -224,7 +224,7 @@ struct redisCommand redisCommandTable[] = {
     {"scan",scanCommand,-2,"rR",0,NULL,0,0,0,0,0},
     {"dbsize",dbsizeCommand,1,"rF",0,NULL,0,0,0,0,0},
     {"auth",authCommand,2,"rsltF",0,NULL,0,0,0,0,0},
-    {"ping",pingCommand,1,"rtF",0,NULL,0,0,0,0,0},
+    {"ping",pingCommand,-1,"rtF",0,NULL,0,0,0,0,0},
     {"echo",echoCommand,2,"rF",0,NULL,0,0,0,0,0},
     {"save",saveCommand,1,"ars",0,NULL,0,0,0,0,0},
     {"bgsave",bgsaveCommand,1,"ar",0,NULL,0,0,0,0,0},
@@ -266,8 +266,8 @@ struct redisCommand redisCommandTable[] = {
     {"readonly",readonlyCommand,1,"rF",0,NULL,0,0,0,0,0},
     {"readwrite",readwriteCommand,1,"rF",0,NULL,0,0,0,0,0},
     {"dump",dumpCommand,2,"ar",0,NULL,1,1,1,0,0},
-    {"object",objectCommand,-2,"r",0,NULL,2,2,2,0,0},
-    {"client",clientCommand,-2,"ar",0,NULL,0,0,0,0,0},
+    {"object",objectCommand,3,"r",0,NULL,2,2,2,0,0},
+    {"client",clientCommand,-2,"ars",0,NULL,0,0,0,0,0},
     {"eval",evalCommand,-3,"s",0,evalGetKeys,0,0,0,0,0},
     {"evalsha",evalShaCommand,-3,"s",0,evalGetKeys,0,0,0,0,0},
     {"slowlog",slowlogCommand,-2,"r",0,NULL,0,0,0,0,0},
@@ -280,7 +280,7 @@ struct redisCommand redisCommandTable[] = {
     {"command",commandCommand,0,"rlt",0,NULL,0,0,0,0,0},
     {"pfselftest",pfselftestCommand,1,"r",0,NULL,0,0,0,0,0},
     {"pfadd",pfaddCommand,-2,"wmF",0,NULL,1,1,1,0,0},
-    {"pfcount",pfcountCommand,-2,"w",0,NULL,1,1,1,0,0},
+    {"pfcount",pfcountCommand,-2,"r",0,NULL,1,1,1,0,0},
     {"pfmerge",pfmergeCommand,-2,"wm",0,NULL,1,-1,1,0,0},
     {"pfdebug",pfdebugCommand,-3,"w",0,NULL,0,0,0,0,0},
     {"latency",latencyCommand,-2,"arslt",0,NULL,0,0,0,0,0}
@@ -408,7 +408,7 @@ void exitFromChild(int retcode) {
 /*====================== Hash table type implementation  ==================== */
 
 /* This is a hash table type that uses the SDS dynamic strings library as
- * keys and radis objects as values (objects can hold SDS strings,
+ * keys and redis objects as values (objects can hold SDS strings,
  * lists, sets). */
 
 void dictVanillaFree(void *privdata, void *val)
@@ -760,8 +760,8 @@ void activeExpireCycle(int type) {
     static int timelimit_exit = 0;      /* Time limit hit in previous call? */
     static long long last_fast_cycle = 0; /* When last fast cycle ran. */
 
-    unsigned int j, iteration = 0;
-    unsigned int dbs_per_call = REDIS_DBCRON_DBS_PER_CALL;
+    int j, iteration = 0;
+    int dbs_per_call = REDIS_DBCRON_DBS_PER_CALL;
     long long start = ustime(), timelimit;
 
     if (type == ACTIVE_EXPIRE_CYCLE_FAST) {
@@ -907,8 +907,7 @@ int clientsCronHandleTimeout(redisClient *c) {
         !(c->flags & REDIS_SLAVE) &&    /* no timeout for slaves */
         !(c->flags & REDIS_MASTER) &&   /* no timeout for masters */
         !(c->flags & REDIS_BLOCKED) &&  /* no timeout for BLPOP */
-        dictSize(c->pubsub_channels) == 0 && /* no timeout for pubsub */
-        listLength(c->pubsub_patterns) == 0 &&
+        !(c->flags & REDIS_PUBSUB) &&   /* no timeout for Pub/Sub clients */
         (now - c->lastinteraction > server.maxidletime))
     {
         redisLog(REDIS_VERBOSE,"Closing idle client");
@@ -1001,8 +1000,8 @@ void databasesCron(void) {
          * cron loop iteration. */
         static unsigned int resize_db = 0;
         static unsigned int rehash_db = 0;
-        unsigned int dbs_per_call = REDIS_DBCRON_DBS_PER_CALL;
-        unsigned int j;
+        int dbs_per_call = REDIS_DBCRON_DBS_PER_CALL;
+        int j;
 
         /* Don't test more DBs than we have. */
         if (dbs_per_call > server.dbnum) dbs_per_call = server.dbnum;
@@ -1377,7 +1376,7 @@ void createSharedObjects(void) {
     shared.maxstring = createStringObject("maxstring",9);
 }
 
-void initServerConfig() {
+void initServerConfig(void) {
     int j;
 
     getRandomHexChars(server.runid,REDIS_RUN_ID_SIZE);
@@ -1421,6 +1420,7 @@ void initServerConfig() {
     server.aof_selected_db = -1; /* Make sure the first time will not match */
     server.aof_flush_postponed_start = 0;
     server.aof_rewrite_incremental_fsync = REDIS_DEFAULT_AOF_REWRITE_INCREMENTAL_FSYNC;
+    server.aof_load_truncated = REDIS_DEFAULT_AOF_LOAD_TRUNCATED;
     server.pidfile = zstrdup(REDIS_DEFAULT_PID_FILE);
     server.rdb_filename = zstrdup(REDIS_DEFAULT_RDB_FILENAME);
     server.aof_filename = zstrdup(REDIS_DEFAULT_AOF_FILENAME);
@@ -1452,6 +1452,7 @@ void initServerConfig() {
     server.cluster_node_timeout = REDIS_CLUSTER_DEFAULT_NODE_TIMEOUT;
     server.cluster_migration_barrier = REDIS_CLUSTER_DEFAULT_MIGRATION_BARRIER;
     server.cluster_slave_validity_factor = REDIS_CLUSTER_DEFAULT_SLAVE_VALIDITY;
+    server.cluster_require_full_coverage = REDIS_CLUSTER_DEFAULT_REQUIRE_FULL_COVERAGE;
     server.cluster_configfile = zstrdup(REDIS_DEFAULT_CLUSTER_CONFIG_FILE);
     server.lua_caller = NULL;
     server.lua_time_limit = REDIS_LUA_TIME_LIMIT;
@@ -1480,6 +1481,8 @@ void initServerConfig() {
     server.repl_slave_ro = REDIS_DEFAULT_SLAVE_READ_ONLY;
     server.repl_down_since = 0; /* Never connected, repl is down since EVER. */
     server.repl_disable_tcp_nodelay = REDIS_DEFAULT_REPL_DISABLE_TCP_NODELAY;
+    server.repl_diskless_sync = REDIS_DEFAULT_REPL_DISKLESS_SYNC;
+    server.repl_diskless_sync_delay = REDIS_DEFAULT_REPL_DISKLESS_SYNC_DELAY;
     server.slave_priority = REDIS_DEFAULT_SLAVE_PRIORITY;
     server.master_repl_offset = 0;
 
@@ -1558,7 +1561,7 @@ void adjustOpenFilesLimit(void) {
              * to the higher value supported less than maxfiles. */
             f = maxfiles;
             while(f > oldlimit) {
-                int decr_step = 16;
+                rlim_t decr_step = 16;
 
                 limit.rlim_cur = f;
                 limit.rlim_max = f;
@@ -1697,7 +1700,7 @@ void resetServerStats(void) {
     server.ops_sec_last_sample_ops = 0;
 }
 
-void initServer() {
+void initServer(void) {
     int j;
 
     signal(SIGHUP, SIG_IGN);
@@ -1768,6 +1771,7 @@ void initServer() {
     server.cronloops = 0;
     server.rdb_child_pid = -1;
     server.aof_child_pid = -1;
+    server.rdb_child_type = REDIS_RDB_CHILD_TYPE_NONE;
     aofRewriteBufferReset();
     server.aof_buf = sdsempty();
     server.lastsave = time(NULL); /* At startup we consider the DB saved. */
@@ -1978,7 +1982,7 @@ void alsoPropagate(struct redisCommand *cmd, int dbid, robj **argv, int argc,
 }
 
 /* It is possible to call the function forceCommandPropagation() inside a
- * Redis command implementaiton in order to to force the propagation of a
+ * Redis command implementation in order to to force the propagation of a
  * specific command execution into AOF / Replication. */
 void forceCommandPropagation(redisClient *c, int flags) {
     if (flags & REDIS_PROPAGATE_REPL) c->flags |= REDIS_FORCE_REPL;
@@ -2007,6 +2011,7 @@ void call(redisClient *c, int flags) {
     c->cmd->proc(c);
     duration = ustime()-start;
     dirty = server.dirty-dirty;
+    if (dirty < 0) dirty = 0;
 
     /* When EVAL is called loading the AOF we don't want commands called
      * from Lua to go into the slowlog or to populate statistics. */
@@ -2211,8 +2216,8 @@ int processCommand(redisClient *c) {
     }
 
     /* Only allow SUBSCRIBE and UNSUBSCRIBE in the context of Pub/Sub */
-    if ((dictSize(c->pubsub_channels) > 0 || listLength(c->pubsub_patterns) > 0)
-        &&
+    if (c->flags & REDIS_PUBSUB &&
+        c->cmd->proc != pingCommand &&
         c->cmd->proc != subscribeCommand &&
         c->cmd->proc != unsubscribeCommand &&
         c->cmd->proc != psubscribeCommand &&
@@ -2360,9 +2365,9 @@ int time_independent_strcmp(char *a, char *b) {
      * a or b are fixed (our password) length, and the difference is only
      * relative to the length of the user provided string, so no information
      * leak is possible in the following two lines of code. */
-    int alen = strlen(a);
-    int blen = strlen(b);
-    int j;
+    unsigned int alen = strlen(a);
+    unsigned int blen = strlen(b);
+    unsigned int j;
     int diff = 0;
 
     /* We can't compare strings longer than our static buffers.
@@ -2399,8 +2404,29 @@ void authCommand(redisClient *c) {
     }
 }
 
+/* The PING command. It works in a different way if the client is in
+ * in Pub/Sub mode. */
 void pingCommand(redisClient *c) {
-    addReply(c,shared.pong);
+    /* The command takes zero or one arguments. */
+    if (c->argc > 2) {
+        addReplyErrorFormat(c,"wrong number of arguments for '%s' command",
+            c->cmd->name);
+        return;
+    }
+
+    if (c->flags & REDIS_PUBSUB) {
+        addReply(c,shared.mbulkhdr[2]);
+        addReplyBulkCBuffer(c,"pong",4);
+        if (c->argc == 1)
+            addReplyBulkCBuffer(c,"",0);
+        else
+            addReplyBulk(c,c->argv[1]);
+    } else {
+        if (c->argc == 1)
+            addReply(c,shared.pong);
+        else
+            addReplyBulk(c,c->argv[1]);
+    }
 }
 
 void echoCommand(redisClient *c) {
@@ -2527,6 +2553,15 @@ void bytesToHuman(char *s, unsigned long long n) {
     } else if (n < (1024LL*1024*1024*1024)) {
         d = (double)n/(1024LL*1024*1024);
         sprintf(s,"%.2fG",d);
+    } else if (n < (1024LL*1024*1024*1024*1024)) {
+        d = (double)n/(1024LL*1024*1024*1024);
+        sprintf(s,"%.2fT",d);
+    } else if (n < (1024LL*1024*1024*1024*1024*1024)) {
+        d = (double)n/(1024LL*1024*1024*1024*1024);
+        sprintf(s,"%.2fP",d);
+    } else {
+        /* Let's hope we never need this */
+        sprintf(s,"%lluB",n);
     }
 }
 
@@ -2542,10 +2577,9 @@ sds genRedisInfoString(char *section) {
     int allsections = 0, defsections = 0;
     int sections = 0;
 
-    if (section) {
-        allsections = strcasecmp(section,"all") == 0;
-        defsections = strcasecmp(section,"default") == 0;
-    }
+    if (section == NULL) section = "default";
+    allsections = strcasecmp(section,"all") == 0;
+    defsections = strcasecmp(section,"default") == 0;
 
     getrusage(RUSAGE_SELF, &self_ru);
     getrusage(RUSAGE_CHILDREN, &c_ru);
@@ -3297,10 +3331,13 @@ int linuxOvercommitMemoryValue(void) {
     return atoi(buf);
 }
 
-void linuxOvercommitMemoryWarning(void) {
+void linuxMemoryWarnings(void) {
     if (linuxOvercommitMemoryValue() == 0) {
         redisLog(REDIS_WARNING,"WARNING overcommit_memory is set to 0! Background save may fail under low memory condition. To fix this issue add 'vm.overcommit_memory = 1' to /etc/sysctl.conf and then reboot or run the command 'sysctl vm.overcommit_memory=1' for this to take effect.");
     }
+    if (THPIsEnabled()) {
+        redisLog(REDIS_WARNING,"WARNING you have Transparent Huge Pages (THP) support enabled in your kernel. This will create latency and memory usage issues with Redis. To fix this issue run the command 'echo never > /sys/kernel/mm/transparent_hugepage/enabled' as root, and add it to your /etc/rc.local in order to retain the setting after a reboot. Redis must be restarted after THP is disabled.");
+    }
 }
 #endif /* __linux__ */
 
@@ -3330,7 +3367,7 @@ void daemonize(void) {
     }
 }
 
-void version() {
+void version(void) {
     printf("Redis server v=%s sha=%s:%d malloc=%s bits=%d build=%llx\n",
         REDIS_VERSION,
         redisGitSHA1(),
@@ -3341,7 +3378,7 @@ void version() {
     exit(0);
 }
 
-void usage() {
+void usage(void) {
     fprintf(stderr,"Usage: ./redis-server [/path/to/redis.conf] [options]\n");
     fprintf(stderr,"       ./redis-server - (read config from stdin)\n");
     fprintf(stderr,"       ./redis-server -v or --version\n");
@@ -3361,10 +3398,11 @@ void usage() {
 void redisAsciiArt(void) {
 #include "asciilogo.h"
     char *buf = zmalloc(1024*16);
-    char *mode = "stand alone";
+    char *mode;
 
     if (server.cluster_enabled) mode = "cluster";
     else if (server.sentinel_mode) mode = "sentinel";
+    else mode = "standalone";
 
     snprintf(buf,1024*16,ascii_logo,
         REDIS_VERSION,
@@ -3378,10 +3416,33 @@ void redisAsciiArt(void) {
     zfree(buf);
 }
 
-static void sigtermHandler(int sig) {
-    REDIS_NOTUSED(sig);
+static void sigShutdownHandler(int sig) {
+    char *msg;
 
-    redisLogFromHandler(REDIS_WARNING,"Received SIGTERM, scheduling shutdown...");
+    switch (sig) {
+    case SIGINT:
+        msg = "Received SIGINT scheduling shutdown...";
+        break;
+    case SIGTERM:
+        msg = "Received SIGTERM scheduling shutdown...";
+        break;
+    default:
+        msg = "Received shutdown signal, scheduling shutdown...";
+    };
+
+    /* SIGINT is often delivered via Ctrl+C in an interactive session.
+     * If we receive the signal the second time, we interpret this as
+     * the user really wanting to quit ASAP without waiting to persist
+     * on disk. */
+    if (server.shutdown_asap && sig == SIGINT) {
+        redisLogFromHandler(REDIS_WARNING, "You insist... exiting now.");
+        rdbRemoveTempFile(getpid());
+        exit(1); /* Exit with an error since this was not a clean shutdown. */
+    } else if (server.loading) {
+        exit(0);
+    }
+
+    redisLogFromHandler(REDIS_WARNING, msg);
     server.shutdown_asap = 1;
 }
 
@@ -3392,8 +3453,9 @@ void setupSignalHandlers(void) {
      * Otherwise, sa_handler is used. */
     sigemptyset(&act.sa_mask);
     act.sa_flags = 0;
-    act.sa_handler = sigtermHandler;
+    act.sa_handler = sigShutdownHandler;
     sigaction(SIGTERM, &act, NULL);
+    sigaction(SIGINT, &act, NULL);
 
 #ifdef HAVE_BACKTRACE
     sigemptyset(&act.sa_mask);
@@ -3524,6 +3586,13 @@ int main(int argc, char **argv) {
             }
             j++;
         }
+        if (server.sentinel_mode && configfile && *configfile == '-') {
+            redisLog(REDIS_WARNING,
+                "Sentinel config from STDIN not allowed.");
+            redisLog(REDIS_WARNING,
+                "Sentinel needs config file on disk to save state.  Exiting...");
+            exit(1);
+        }
         if (configfile) server.configfile = getAbsolutePath(configfile);
         resetServerSaveParams();
         loadServerConfig(configfile,options);
@@ -3541,7 +3610,7 @@ int main(int argc, char **argv) {
         /* Things not needed when running in Sentinel mode. */
         redisLog(REDIS_WARNING,"Server started, Redis version " REDIS_VERSION);
     #ifdef __linux__
-        linuxOvercommitMemoryWarning();
+        linuxMemoryWarnings();
     #endif
         loadDataFromDisk();
         if (server.cluster_enabled) {

src/redis.h

@@ -96,6 +96,7 @@ typedef long long mstime_t; /* millisecond time type. */
 #define REDIS_REPL_TIMEOUT 60
 #define REDIS_REPL_PING_SLAVE_PERIOD 10
 #define REDIS_RUN_ID_SIZE 40
+#define REDIS_EOF_MARK_SIZE 40
 #define REDIS_OPS_SEC_SAMPLES 16
 #define REDIS_DEFAULT_REPL_BACKLOG_SIZE (1024*1024)    /* 1mb */
 #define REDIS_DEFAULT_REPL_BACKLOG_TIME_LIMIT (60*60)  /* 1 hour */
@@ -113,6 +114,8 @@ typedef long long mstime_t; /* millisecond time type. */
 #define REDIS_DEFAULT_RDB_COMPRESSION 1
 #define REDIS_DEFAULT_RDB_CHECKSUM 1
 #define REDIS_DEFAULT_RDB_FILENAME "dump.rdb"
+#define REDIS_DEFAULT_REPL_DISKLESS_SYNC 0
+#define REDIS_DEFAULT_REPL_DISKLESS_SYNC_DELAY 5
 #define REDIS_DEFAULT_SLAVE_SERVE_STALE_DATA 1
 #define REDIS_DEFAULT_SLAVE_READ_ONLY 1
 #define REDIS_DEFAULT_REPL_DISABLE_TCP_NODELAY 0
@@ -120,6 +123,7 @@ typedef long long mstime_t; /* millisecond time type. */
 #define REDIS_DEFAULT_MAXMEMORY_SAMPLES 5
 #define REDIS_DEFAULT_AOF_FILENAME "appendonly.aof"
 #define REDIS_DEFAULT_AOF_NO_FSYNC_ON_REWRITE 0
+#define REDIS_DEFAULT_AOF_LOAD_TRUNCATED 1
 #define REDIS_DEFAULT_ACTIVE_REHASHING 1
 #define REDIS_DEFAULT_AOF_REWRITE_INCREMENTAL_FSYNC 1
 #define REDIS_DEFAULT_MIN_SLAVES_TO_WRITE 0
@@ -241,6 +245,7 @@ typedef long long mstime_t; /* millisecond time type. */
 #define REDIS_FORCE_REPL (1<<15)  /* Force replication of current cmd. */
 #define REDIS_PRE_PSYNC (1<<16)   /* Instance don't understand PSYNC. */
 #define REDIS_READONLY (1<<17)    /* Cluster client is in read-only state. */
+#define REDIS_PUBSUB (1<<18)      /* Client is in Pub/Sub mode. */
 
 /* Client block type (btype field in client structure)
  * if REDIS_BLOCKED flag is set. */
@@ -359,6 +364,11 @@ typedef long long mstime_t; /* millisecond time type. */
 #define REDIS_PROPAGATE_AOF 1
 #define REDIS_PROPAGATE_REPL 2
 
+/* RDB active child save type. */
+#define REDIS_RDB_CHILD_TYPE_NONE 0
+#define REDIS_RDB_CHILD_TYPE_DISK 1     /* RDB is written to disk. */
+#define REDIS_RDB_CHILD_TYPE_SOCKET 2   /* RDB is written to slave socket. */
+
 /* Keyspace changes notification classes. Every class is associated with a
  * character for configuration purposes. */
 #define REDIS_NOTIFY_KEYSPACE (1<<0)    /* K */
@@ -497,7 +507,7 @@ typedef struct readyList {
 } readyList;
 
 /* With multiplexing we need to take per-client state.
- * Clients are taken in a liked list. */
+ * Clients are taken in a linked list. */
 typedef struct redisClient {
     uint64_t id;            /* Client incremental unique ID. */
     int fd;
@@ -522,6 +532,7 @@ typedef struct redisClient {
     int flags;              /* REDIS_SLAVE | REDIS_MONITOR | REDIS_MULTI ... */
     int authenticated;      /* when requirepass is non-NULL */
     int replstate;          /* replication state if this is a slave */
+    int repl_put_online_on_ack; /* Install slave write handler on ACK. */
     int repldbfd;           /* replication DB file descriptor */
     off_t repldboff;        /* replication DB file offset */
     off_t repldbsize;       /* replication DB file size */
@@ -625,6 +636,12 @@ typedef struct redisOpArray {
 
 struct clusterState;
 
+/* AIX defines hz to __hz, we don't use this define and in order to allow
+ * Redis build on AIX we need to undef it. */
+#ifdef _AIX
+#undef hz
+#endif
+
 struct redisServer {
     /* General */
     pid_t pid;                  /* Main process pid. */
@@ -732,6 +749,7 @@ struct redisServer {
     int aof_rewrite_incremental_fsync;/* fsync incrementally while rewriting? */
     int aof_last_write_status;      /* REDIS_OK or REDIS_ERR */
     int aof_last_write_errno;       /* Valid if aof_last_write_status is ERR */
+    int aof_load_truncated;         /* Don't stop on unexpected AOF EOF. */
     /* AOF pipes used to communicate between parent and child during rewrite. */
     int aof_pipe_write_data_to_child;
     int aof_pipe_read_data_from_parent;
@@ -755,8 +773,11 @@ struct redisServer {
     time_t lastbgsave_try;          /* Unix time of last attempted bgsave */
     time_t rdb_save_time_last;      /* Time used by last RDB save run. */
     time_t rdb_save_time_start;     /* Current RDB save start time. */
+    int rdb_child_type;             /* Type of save by active child. */
     int lastbgsave_status;          /* REDIS_OK or REDIS_ERR */
     int stop_writes_on_bgsave_err;  /* Don't allow writes if can't BGSAVE */
+    int rdb_pipe_write_result_to_parent; /* RDB pipes used to return the state */
+    int rdb_pipe_read_result_from_child; /* of each slave in diskless SYNC. */
     /* Propagation of commands in AOF / replication */
     redisOpArray also_propagate;    /* Additional command to propagate. */
     /* Logging */
@@ -781,6 +802,8 @@ struct redisServer {
     int repl_min_slaves_to_write;   /* Min number of slaves to write. */
     int repl_min_slaves_max_lag;    /* Max lag of <count> slaves to write. */
     int repl_good_slaves_count;     /* Number of slaves with lag <= max_lag. */
+    int repl_diskless_sync;         /* Send RDB to slaves sockets directly. */
+    int repl_diskless_sync_delay;   /* Delay to start a diskless repl BGSAVE. */
     /* Replication (slave) */
     char *masterauth;               /* AUTH with this password with master */
     char *masterhost;               /* Hostname of master */
@@ -807,12 +830,12 @@ struct redisServer {
     /* Replication script cache. */
     dict *repl_scriptcache_dict;        /* SHA1 all slaves are aware of. */
     list *repl_scriptcache_fifo;        /* First in, first out LRU eviction. */
-    int repl_scriptcache_size;          /* Max number of elements. */
+    unsigned int repl_scriptcache_size; /* Max number of elements. */
     /* Synchronous replication. */
     list *clients_waiting_acks;         /* Clients waiting in WAIT command. */
     int get_ack_from_slaves;            /* If true we send REPLCONF GETACK. */
     /* Limits */
-    int maxclients;                 /* Max number of simultaneous clients */
+    unsigned int maxclients;            /* Max number of simultaneous clients */
     unsigned long long maxmemory;   /* Max number of memory bytes to use */
     int maxmemory_policy;           /* Policy for key eviction */
     int maxmemory_samples;          /* Pricision of random sampling */
@@ -849,6 +872,8 @@ struct redisServer {
     struct clusterState *cluster;  /* State of the cluster */
     int cluster_migration_barrier; /* Cluster replicas migration barrier. */
     int cluster_slave_validity_factor; /* Slave max data age for failover. */
+    int cluster_require_full_coverage; /* If true, put the cluster down if
+                                          there is at least an uncovered slot. */
     /* Scripting */
     lua_State *lua; /* The Lua interpreter. We use just one for all clients */
     redisClient *lua_client;   /* The "fake client" to query Redis from Lua */
@@ -992,11 +1017,10 @@ void freeClient(redisClient *c);
 void freeClientAsync(redisClient *c);
 void resetClient(redisClient *c);
 void sendReplyToClient(aeEventLoop *el, int fd, void *privdata, int mask);
-void addReply(redisClient *c, robj *obj);
 void *addDeferredMultiBulkLength(redisClient *c);
 void setDeferredMultiBulkLength(redisClient *c, void *node, long length);
-void addReplySds(redisClient *c, sds s);
 void processInputBuffer(redisClient *c);
+void acceptHandler(aeEventLoop *el, int fd, void *privdata, int mask);
 void acceptTcpHandler(aeEventLoop *el, int fd, void *privdata, int mask);
 void acceptUnixHandler(aeEventLoop *el, int fd, void *privdata, int mask);
 void readQueryFromClient(aeEventLoop *el, int fd, void *privdata, int mask);
@@ -1004,7 +1028,6 @@ void addReplyBulk(redisClient *c, robj *obj);
 void addReplyBulkCString(redisClient *c, char *s);
 void addReplyBulkCBuffer(redisClient *c, void *p, size_t len);
 void addReplyBulkLongLong(redisClient *c, long long ll);
-void acceptHandler(aeEventLoop *el, int fd, void *privdata, int mask);
 void addReply(redisClient *c, robj *obj);
 void addReplySds(redisClient *c, sds s);
 void addReplyError(redisClient *c, char *err);
@@ -1093,7 +1116,7 @@ robj *tryObjectEncoding(robj *o);
 robj *getDecodedObject(robj *o);
 size_t stringObjectLen(robj *o);
 robj *createStringObjectFromLongLong(long long value);
-robj *createStringObjectFromLongDouble(long double value);
+robj *createStringObjectFromLongDouble(long double value, int humanfriendly);
 robj *createListObject(void);
 robj *createZiplistObject(void);
 robj *createSetObject(void);
@@ -1123,7 +1146,7 @@ ssize_t syncReadLine(int fd, char *ptr, ssize_t size, long long timeout);
 /* Replication */
 void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc);
 void replicationFeedMonitors(redisClient *c, list *monitors, int dictid, robj **argv, int argc);
-void updateSlavesWaitingBgsave(int bgsaveerr);
+void updateSlavesWaitingBgsave(int bgsaveerr, int type);
 void replicationCron(void);
 void replicationHandleMasterDisconnection(void);
 void replicationCacheMaster(redisClient *c);
@@ -1140,6 +1163,7 @@ void unblockClientWaitingReplicas(redisClient *c);
 int replicationCountAcksByOffset(long long offset);
 void replicationSendNewlineToMaster(void);
 long long replicationGetSlaveOffset(void);
+char *replicationGetSlaveName(redisClient *c);
 
 /* Generic persistence functions */
 void startLoading(FILE *fp);
@@ -1209,7 +1233,7 @@ void redisLog(int level, const char *fmt, ...);
 #endif
 void redisLogRaw(int level, const char *msg);
 void redisLogFromHandler(int level, const char *msg);
-void usage();
+void usage(void);
 void updateDictResizePolicy(void);
 int htNeedsResize(dict *dict);
 void oom(const char *msg);
@@ -1269,7 +1293,7 @@ sds keyspaceEventsFlagsToString(int flags);
 /* Configuration */
 void loadServerConfig(char *filename, char *options);
 void appendServerSaveParams(time_t seconds, int changes);
-void resetServerSaveParams();
+void resetServerSaveParams(void);
 struct rewriteConfigState; /* Forward declaration to export API. */
 void rewriteConfigRewriteLine(struct rewriteConfigState *state, char *option, sds line, int force);
 int rewriteConfig(char *path);

src/replication.c

@@ -40,6 +40,31 @@
 void replicationDiscardCachedMaster(void);
 void replicationResurrectCachedMaster(int newfd);
 void replicationSendAck(void);
+void putSlaveOnline(redisClient *slave);
+
+/* --------------------------- Utility functions ---------------------------- */
+
+/* Return the pointer to a string representing the slave ip:listening_port
+ * pair. Mostly useful for logging, since we want to log a slave using its
+ * IP address and it's listening port which is more clear for the user, for
+ * example: "Closing connection with slave 10.1.2.3:6380". */
+char *replicationGetSlaveName(redisClient *c) {
+    static char buf[REDIS_PEER_ID_LEN];
+    char ip[REDIS_IP_STR_LEN];
+
+    ip[0] = '\0';
+    buf[0] = '\0';
+    if (anetPeerToString(c->fd,ip,sizeof(ip),NULL) != -1) {
+        if (c->slave_listening_port)
+            snprintf(buf,sizeof(buf),"%s:%d",ip,c->slave_listening_port);
+        else
+            snprintf(buf,sizeof(buf),"%s:<unknown-slave-port>",ip);
+    } else {
+        snprintf(buf,sizeof(buf),"client id #%llu",
+            (unsigned long long) c->id);
+    }
+    return buf;
+}
 
 /* ---------------------------------- MASTER -------------------------------- */
 
@@ -82,7 +107,7 @@ void resizeReplicationBacklog(long long newsize) {
         server.repl_backlog = zmalloc(server.repl_backlog_size);
         server.repl_backlog_histlen = 0;
         server.repl_backlog_idx = 0;
-        /* Next byte we have is... the next since the buffer is emtpy. */
+        /* Next byte we have is... the next since the buffer is empty. */
         server.repl_backlog_off = server.master_repl_offset+1;
     }
 }
@@ -200,7 +225,7 @@ void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc) {
 
             /* We need to feed the buffer with the object as a bulk reply
              * not just as a plain string, so create the $..CRLF payload len
-             * ad add the final CRLF */
+             * and add the final CRLF */
             aux[0] = '$';
             len = ll2string(aux+1,sizeof(aux)-1,objlen);
             aux[len+1] = '\r';
@@ -212,7 +237,7 @@ void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc) {
     }
 
     /* Write the command to every slave. */
-    listRewind(slaves,&li);
+    listRewind(server.slaves,&li);
     while((ln = listNext(&li))) {
         redisClient *slave = ln->value;
 
@@ -345,7 +370,8 @@ int masterTryPartialResynchronization(redisClient *c) {
                 "Runid mismatch (Client asked for runid '%s', my runid is '%s')",
                 master_runid, server.runid);
         } else {
-            redisLog(REDIS_NOTICE,"Full resync requested by slave.");
+            redisLog(REDIS_NOTICE,"Full resync requested by slave %s",
+                replicationGetSlaveName(c));
         }
         goto need_full_resync;
     }
@@ -358,10 +384,10 @@ int masterTryPartialResynchronization(redisClient *c) {
         psync_offset > (server.repl_backlog_off + server.repl_backlog_histlen))
     {
         redisLog(REDIS_NOTICE,
-            "Unable to partial resync with the slave for lack of backlog (Slave request was: %lld).", psync_offset);
+            "Unable to partial resync with slave %s for lack of backlog (Slave request was: %lld).", replicationGetSlaveName(c), psync_offset);
         if (psync_offset > server.master_repl_offset) {
             redisLog(REDIS_WARNING,
-                "Warning: slave tried to PSYNC with an offset that is greater than the master replication offset.");
+                "Warning: slave %s tried to PSYNC with an offset that is greater than the master replication offset.", replicationGetSlaveName(c));
         }
         goto need_full_resync;
     }
@@ -373,10 +399,11 @@ int masterTryPartialResynchronization(redisClient *c) {
     c->flags |= REDIS_SLAVE;
     c->replstate = REDIS_REPL_ONLINE;
     c->repl_ack_time = server.unixtime;
+    c->repl_put_online_on_ack = 0;
     listAddNodeTail(server.slaves,c);
     /* We can't use the connection buffers since they are used to accumulate
      * new commands at this stage. But we are sure the socket send buffer is
-     * emtpy so this write will never fail actually. */
+     * empty so this write will never fail actually. */
     buflen = snprintf(buf,sizeof(buf),"+CONTINUE\r\n");
     if (write(c->fd,buf,buflen) != buflen) {
         freeClientAsync(c);
@@ -384,7 +411,9 @@ int masterTryPartialResynchronization(redisClient *c) {
     }
     psync_len = addReplyReplicationBacklog(c,psync_offset);
     redisLog(REDIS_NOTICE,
-        "Partial resynchronization request accepted. Sending %lld bytes of backlog starting from offset %lld.", psync_len, psync_offset);
+        "Partial resynchronization request from %s accepted. Sending %lld bytes of backlog starting from offset %lld.",
+            replicationGetSlaveName(c),
+            psync_len, psync_offset);
     /* Note that we don't need to set the selected DB at server.slaveseldb
      * to -1 to force the master to emit SELECT, since the slave already
      * has this state from the previous connection with the master. */
@@ -408,7 +437,29 @@ need_full_resync:
     return REDIS_ERR;
 }
 
-/* SYNC ad PSYNC command implemenation. */
+/* Start a BGSAVE for replication goals, which is, selecting the disk or
+ * socket target depending on the configuration, and making sure that
+ * the script cache is flushed before to start.
+ *
+ * Returns REDIS_OK on success or REDIS_ERR otherwise. */
+int startBgsaveForReplication(void) {
+    int retval;
+
+    redisLog(REDIS_NOTICE,"Starting BGSAVE for SYNC with target: %s",
+        server.repl_diskless_sync ? "slaves sockets" : "disk");
+
+    if (server.repl_diskless_sync)
+        retval = rdbSaveToSlavesSockets();
+    else
+        retval = rdbSaveBackground(server.rdb_filename);
+
+    /* Flush the script cache, since we need that slave differences are
+     * accumulated without requiring slaves to match our cached scripts. */
+    if (retval == REDIS_OK) replicationScriptCacheFlush();
+    return retval;
+}
+
+/* SYNC and PSYNC command implemenation. */
 void syncCommand(redisClient *c) {
     /* ignore SYNC if already slave or in monitor mode */
     if (c->flags & REDIS_SLAVE) return;
@@ -429,7 +480,8 @@ void syncCommand(redisClient *c) {
         return;
     }
 
-    redisLog(REDIS_NOTICE,"Slave asks for synchronization");
+    redisLog(REDIS_NOTICE,"Slave %s asks for synchronization",
+        replicationGetSlaveName(c));
 
     /* Try a partial resynchronization if this is a PSYNC command.
      * If it fails, we continue with usual full resynchronization, however
@@ -465,10 +517,12 @@ void syncCommand(redisClient *c) {
 
     /* Here we need to check if there is a background saving operation
      * in progress, or if it is required to start one */
-    if (server.rdb_child_pid != -1) {
+    if (server.rdb_child_pid != -1 &&
+        server.rdb_child_type == REDIS_RDB_CHILD_TYPE_DISK)
+    {
         /* Ok a background save is in progress. Let's check if it is a good
          * one for replication, i.e. if there is another slave that is
-         * registering differences since the server forked to save */
+         * registering differences since the server forked to save. */
         redisClient *slave;
         listNode *ln;
         listIter li;
@@ -486,21 +540,35 @@ void syncCommand(redisClient *c) {
             redisLog(REDIS_NOTICE,"Waiting for end of BGSAVE for SYNC");
         } else {
             /* No way, we need to wait for the next BGSAVE in order to
-             * register differences */
+             * register differences. */
             c->replstate = REDIS_REPL_WAIT_BGSAVE_START;
             redisLog(REDIS_NOTICE,"Waiting for next BGSAVE for SYNC");
         }
+    } else if (server.rdb_child_pid != -1 &&
+               server.rdb_child_type == REDIS_RDB_CHILD_TYPE_SOCKET)
+    {
+        /* There is an RDB child process but it is writing directly to
+         * children sockets. We need to wait for the next BGSAVE
+         * in order to synchronize. */
+        c->replstate = REDIS_REPL_WAIT_BGSAVE_START;
+        redisLog(REDIS_NOTICE,"Waiting for next BGSAVE for SYNC");
     } else {
-        /* Ok we don't have a BGSAVE in progress, let's start one */
-        redisLog(REDIS_NOTICE,"Starting BGSAVE for SYNC");
-        if (rdbSaveBackground(server.rdb_filename) != REDIS_OK) {
-            redisLog(REDIS_NOTICE,"Replication failed, can't BGSAVE");
-            addReplyError(c,"Unable to perform background save");
-            return;
+        if (server.repl_diskless_sync) {
+            /* Diskless replication RDB child is created inside
+             * replicationCron() since we want to delay its start a
+             * few seconds to wait for more slaves to arrive. */
+            c->replstate = REDIS_REPL_WAIT_BGSAVE_START;
+            if (server.repl_diskless_sync_delay)
+                redisLog(REDIS_NOTICE,"Delay next BGSAVE for SYNC");
+        } else {
+            /* Ok we don't have a BGSAVE in progress, let's start one. */
+            if (startBgsaveForReplication() != REDIS_OK) {
+                redisLog(REDIS_NOTICE,"Replication failed, can't BGSAVE");
+                addReplyError(c,"Unable to perform background save");
+                return;
+            }
+            c->replstate = REDIS_REPL_WAIT_BGSAVE_END;
         }
-        c->replstate = REDIS_REPL_WAIT_BGSAVE_END;
-        /* Flush the script cache for the new slave. */
-        replicationScriptCacheFlush();
     }
 
     if (server.repl_disable_tcp_nodelay)
@@ -557,6 +625,11 @@ void replconfCommand(redisClient *c) {
             if (offset > c->repl_ack_off)
                 c->repl_ack_off = offset;
             c->repl_ack_time = server.unixtime;
+            /* If this was a diskless replication, we need to really put
+             * the slave online when the first ACK is received (which
+             * confirms slave is online and ready to get more data). */
+            if (c->repl_put_online_on_ack && c->replstate == REDIS_REPL_ONLINE)
+                putSlaveOnline(c);
             /* Note: this command does not reply anything! */
             return;
         } else if (!strcasecmp(c->argv[j]->ptr,"getack")) {
@@ -573,6 +646,32 @@ void replconfCommand(redisClient *c) {
     addReply(c,shared.ok);
 }
 
+/* This function puts a slave in the online state, and should be called just
+ * after a slave received the RDB file for the initial synchronization, and
+ * we are finally ready to send the incremental stream of commands.
+ *
+ * It does a few things:
+ *
+ * 1) Put the slave in ONLINE state.
+ * 2) Make sure the writable event is re-installed, since calling the SYNC
+ *    command disables it, so that we can accumulate output buffer without
+ *    sending it to the slave.
+ * 3) Update the count of good slaves. */
+void putSlaveOnline(redisClient *slave) {
+    slave->replstate = REDIS_REPL_ONLINE;
+    slave->repl_put_online_on_ack = 0;
+    slave->repl_ack_time = server.unixtime;
+    if (aeCreateFileEvent(server.el, slave->fd, AE_WRITABLE,
+        sendReplyToClient, slave) == AE_ERR) {
+        redisLog(REDIS_WARNING,"Unable to register writable event for slave bulk transfer: %s", strerror(errno));
+        freeClient(slave);
+        return;
+    }
+    refreshGoodSlavesCount();
+    redisLog(REDIS_NOTICE,"Synchronization with slave %s succeeded",
+        replicationGetSlaveName(slave));
+}
+
 void sendBulkToSlave(aeEventLoop *el, int fd, void *privdata, int mask) {
     redisClient *slave = privdata;
     REDIS_NOTUSED(el);
@@ -623,26 +722,25 @@ void sendBulkToSlave(aeEventLoop *el, int fd, void *privdata, int mask) {
         close(slave->repldbfd);
         slave->repldbfd = -1;
         aeDeleteFileEvent(server.el,slave->fd,AE_WRITABLE);
-        slave->replstate = REDIS_REPL_ONLINE;
-        slave->repl_ack_time = server.unixtime;
-        if (aeCreateFileEvent(server.el, slave->fd, AE_WRITABLE,
-            sendReplyToClient, slave) == AE_ERR) {
-            redisLog(REDIS_WARNING,"Unable to register writable event for slave bulk transfer: %s", strerror(errno));
-            freeClient(slave);
-            return;
-        }
-        refreshGoodSlavesCount();
-        redisLog(REDIS_NOTICE,"Synchronization with slave succeeded");
+        putSlaveOnline(slave);
     }
 }
 
-/* This function is called at the end of every background saving.
- * The argument bgsaveerr is REDIS_OK if the background saving succeeded
- * otherwise REDIS_ERR is passed to the function.
+/* This function is called at the end of every background saving,
+ * or when the replication RDB transfer strategy is modified from
+ * disk to socket or the other way around.
  *
  * The goal of this function is to handle slaves waiting for a successful
- * background saving in order to perform non-blocking synchronization. */
-void updateSlavesWaitingBgsave(int bgsaveerr) {
+ * background saving in order to perform non-blocking synchronization, and
+ * to schedule a new BGSAVE if there are slaves that attached while a
+ * BGSAVE was in progress, but it was not a good one for replication (no
+ * other slave was accumulating differences).
+ *
+ * The argument bgsaveerr is REDIS_OK if the background saving succeeded
+ * otherwise REDIS_ERR is passed to the function.
+ * The 'type' argument is the type of the child that terminated
+ * (if it had a disk or socket target). */
+void updateSlavesWaitingBgsave(int bgsaveerr, int type) {
     listNode *ln;
     int startbgsave = 0;
     listIter li;
@@ -657,37 +755,50 @@ void updateSlavesWaitingBgsave(int bgsaveerr) {
         } else if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_END) {
             struct redis_stat buf;
 
-            if (bgsaveerr != REDIS_OK) {
-                freeClient(slave);
-                redisLog(REDIS_WARNING,"SYNC failed. BGSAVE child returned an error");
-                continue;
-            }
-            if ((slave->repldbfd = open(server.rdb_filename,O_RDONLY)) == -1 ||
-                redis_fstat(slave->repldbfd,&buf) == -1) {
-                freeClient(slave);
-                redisLog(REDIS_WARNING,"SYNC failed. Can't open/stat DB after BGSAVE: %s", strerror(errno));
-                continue;
-            }
-            slave->repldboff = 0;
-            slave->repldbsize = buf.st_size;
-            slave->replstate = REDIS_REPL_SEND_BULK;
-            slave->replpreamble = sdscatprintf(sdsempty(),"$%lld\r\n",
-                (unsigned long long) slave->repldbsize);
+            /* If this was an RDB on disk save, we have to prepare to send
+             * the RDB from disk to the slave socket. Otherwise if this was
+             * already an RDB -> Slaves socket transfer, used in the case of
+             * diskless replication, our work is trivial, we can just put
+             * the slave online. */
+            if (type == REDIS_RDB_CHILD_TYPE_SOCKET) {
+                redisLog(REDIS_NOTICE,
+                    "Streamed RDB transfer with slave %s succeeded (socket). Waiting for REPLCONF ACK from slave to enable streaming",
+                        replicationGetSlaveName(slave));
+                /* Note: we wait for a REPLCONF ACK message from slave in
+                 * order to really put it online (install the write handler
+                 * so that the accumulated data can be transfered). However
+                 * we change the replication state ASAP, since our slave
+                 * is technically online now. */
+                slave->replstate = REDIS_REPL_ONLINE;
+                slave->repl_put_online_on_ack = 1;
+            } else {
+                if (bgsaveerr != REDIS_OK) {
+                    freeClient(slave);
+                    redisLog(REDIS_WARNING,"SYNC failed. BGSAVE child returned an error");
+                    continue;
+                }
+                if ((slave->repldbfd = open(server.rdb_filename,O_RDONLY)) == -1 ||
+                    redis_fstat(slave->repldbfd,&buf) == -1) {
+                    freeClient(slave);
+                    redisLog(REDIS_WARNING,"SYNC failed. Can't open/stat DB after BGSAVE: %s", strerror(errno));
+                    continue;
+                }
+                slave->repldboff = 0;
+                slave->repldbsize = buf.st_size;
+                slave->replstate = REDIS_REPL_SEND_BULK;
+                slave->replpreamble = sdscatprintf(sdsempty(),"$%lld\r\n",
+                    (unsigned long long) slave->repldbsize);
 
-            aeDeleteFileEvent(server.el,slave->fd,AE_WRITABLE);
-            if (aeCreateFileEvent(server.el, slave->fd, AE_WRITABLE, sendBulkToSlave, slave) == AE_ERR) {
-                freeClient(slave);
-                continue;
+                aeDeleteFileEvent(server.el,slave->fd,AE_WRITABLE);
+                if (aeCreateFileEvent(server.el, slave->fd, AE_WRITABLE, sendBulkToSlave, slave) == AE_ERR) {
+                    freeClient(slave);
+                    continue;
+                }
             }
         }
     }
     if (startbgsave) {
-        /* Since we are starting a new background save for one or more slaves,
-         * we flush the Replication Script Cache to use EVAL to propagate every
-         * new EVALSHA for the first time, since all the new slaves don't know
-         * about previous scripts. */
-        replicationScriptCacheFlush();
-        if (rdbSaveBackground(server.rdb_filename) != REDIS_OK) {
+        if (startBgsaveForReplication() != REDIS_OK) {
             listIter li;
 
             listRewind(server.slaves,&li);
@@ -751,6 +862,12 @@ void readSyncBulkPayload(aeEventLoop *el, int fd, void *privdata, int mask) {
     REDIS_NOTUSED(privdata);
     REDIS_NOTUSED(mask);
 
+    /* Static vars used to hold the EOF mark, and the last bytes received
+     * form the server: when they match, we reached the end of the transfer. */
+    static char eofmark[REDIS_RUN_ID_SIZE];
+    static char lastbytes[REDIS_RUN_ID_SIZE];
+    static int usemark = 0;
+
     /* If repl_transfer_size == -1 we still have to read the bulk length
      * from the master reply. */
     if (server.repl_transfer_size == -1) {
@@ -776,16 +893,44 @@ void readSyncBulkPayload(aeEventLoop *el, int fd, void *privdata, int mask) {
             redisLog(REDIS_WARNING,"Bad protocol from MASTER, the first byte is not '$' (we received '%s'), are you sure the host and port are right?", buf);
             goto error;
         }
-        server.repl_transfer_size = strtol(buf+1,NULL,10);
-        redisLog(REDIS_NOTICE,
-            "MASTER <-> SLAVE sync: receiving %lld bytes from master",
-            (long long) server.repl_transfer_size);
+
+        /* There are two possible forms for the bulk payload. One is the
+         * usual $<count> bulk format. The other is used for diskless transfers
+         * when the master does not know beforehand the size of the file to
+         * transfer. In the latter case, the following format is used:
+         *
+         * $EOF:<40 bytes delimiter>
+         *
+         * At the end of the file the announced delimiter is transmitted. The
+         * delimiter is long and random enough that the probability of a
+         * collision with the actual file content can be ignored. */
+        if (strncmp(buf+1,"EOF:",4) == 0 && strlen(buf+5) >= REDIS_RUN_ID_SIZE) {
+            usemark = 1;
+            memcpy(eofmark,buf+5,REDIS_RUN_ID_SIZE);
+            memset(lastbytes,0,REDIS_RUN_ID_SIZE);
+            /* Set any repl_transfer_size to avoid entering this code path
+             * at the next call. */
+            server.repl_transfer_size = 0;
+            redisLog(REDIS_NOTICE,
+                "MASTER <-> SLAVE sync: receiving streamed RDB from master");
+        } else {
+            usemark = 0;
+            server.repl_transfer_size = strtol(buf+1,NULL,10);
+            redisLog(REDIS_NOTICE,
+                "MASTER <-> SLAVE sync: receiving %lld bytes from master",
+                (long long) server.repl_transfer_size);
+        }
         return;
     }
 
     /* Read bulk data */
-    left = server.repl_transfer_size - server.repl_transfer_read;
-    readlen = (left < (signed)sizeof(buf)) ? left : (signed)sizeof(buf);
+    if (usemark) {
+        readlen = sizeof(buf);
+    } else {
+        left = server.repl_transfer_size - server.repl_transfer_read;
+        readlen = (left < (signed)sizeof(buf)) ? left : (signed)sizeof(buf);
+    }
+
     nread = read(fd,buf,readlen);
     if (nread <= 0) {
         redisLog(REDIS_WARNING,"I/O error trying to sync with MASTER: %s",
@@ -793,6 +938,23 @@ void readSyncBulkPayload(aeEventLoop *el, int fd, void *privdata, int mask) {
         replicationAbortSyncTransfer();
         return;
     }
+
+    /* When a mark is used, we want to detect EOF asap in order to avoid
+     * writing the EOF mark into the file... */
+    int eof_reached = 0;
+
+    if (usemark) {
+        /* Update the last bytes array, and check if it matches our delimiter.*/
+        if (nread >= REDIS_RUN_ID_SIZE) {
+            memcpy(lastbytes,buf+nread-REDIS_RUN_ID_SIZE,REDIS_RUN_ID_SIZE);
+        } else {
+            int rem = REDIS_RUN_ID_SIZE-nread;
+            memmove(lastbytes,lastbytes+nread,rem);
+            memcpy(lastbytes+rem,buf,nread);
+        }
+        if (memcmp(lastbytes,eofmark,REDIS_RUN_ID_SIZE) == 0) eof_reached = 1;
+    }
+
     server.repl_transfer_lastio = server.unixtime;
     if (write(server.repl_transfer_fd,buf,nread) != nread) {
         redisLog(REDIS_WARNING,"Write error or short write writing to the DB dump file needed for MASTER <-> SLAVE synchronization: %s", strerror(errno));
@@ -800,6 +962,16 @@ void readSyncBulkPayload(aeEventLoop *el, int fd, void *privdata, int mask) {
     }
     server.repl_transfer_read += nread;
 
+    /* Delete the last 40 bytes from the file if we reached EOF. */
+    if (usemark && eof_reached) {
+        if (ftruncate(server.repl_transfer_fd,
+            server.repl_transfer_read - REDIS_RUN_ID_SIZE) == -1)
+        {
+            redisLog(REDIS_WARNING,"Error truncating the RDB file received from the master for SYNC: %s", strerror(errno));
+            goto error;
+        }
+    }
+
     /* Sync data on disk from time to time, otherwise at the end of the transfer
      * we may suffer a big delay as the memory buffers are copied into the
      * actual disk. */
@@ -814,7 +986,12 @@ void readSyncBulkPayload(aeEventLoop *el, int fd, void *privdata, int mask) {
     }
 
     /* Check if the transfer is now complete */
-    if (server.repl_transfer_read == server.repl_transfer_size) {
+    if (!usemark) {
+        if (server.repl_transfer_read == server.repl_transfer_size)
+            eof_reached = 1;
+    }
+
+    if (eof_reached) {
         if (rename(server.repl_transfer_tmpfile,server.rdb_filename) == -1) {
             redisLog(REDIS_WARNING,"Failed trying to rename the temp DB into dump.rdb in MASTER <-> SLAVE synchronization: %s", strerror(errno));
             replicationAbortSyncTransfer();
@@ -1195,7 +1372,8 @@ error:
 int connectWithMaster(void) {
     int fd;
 
-    fd = anetTcpNonBlockConnect(NULL,server.masterhost,server.masterport);
+    fd = anetTcpNonBlockBindConnect(NULL,
+        server.masterhost,server.masterport,REDIS_BIND_ADDR);
     if (fd == -1) {
         redisLog(REDIS_WARNING,"Unable to connect to MASTER: %s",
             strerror(errno));
@@ -1460,7 +1638,7 @@ void replicationDiscardCachedMaster(void) {
 /* Turn the cached master into the current master, using the file descriptor
  * passed as argument as the socket for the new master.
  *
- * This funciton is called when successfully setup a partial resynchronization
+ * This function is called when successfully setup a partial resynchronization
  * so the stream of data that we'll receive will start from were this
  * master left. */
 void replicationResurrectCachedMaster(int newfd) {
@@ -1750,7 +1928,7 @@ long long replicationGetSlaveOffset(void) {
 
 /* --------------------------- REPLICATION CRON  ---------------------------- */
 
-/* Replication cron funciton, called 1 time per second. */
+/* Replication cron function, called 1 time per second. */
 void replicationCron(void) {
     /* Non blocking connection timeout? */
     if (server.masterhost &&
@@ -1817,7 +1995,9 @@ void replicationCron(void) {
             redisClient *slave = ln->value;
 
             if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_START ||
-                slave->replstate == REDIS_REPL_WAIT_BGSAVE_END) {
+                (slave->replstate == REDIS_REPL_WAIT_BGSAVE_END &&
+                 server.rdb_child_type != REDIS_RDB_CHILD_TYPE_SOCKET))
+            {
                 if (write(slave->fd, "\n", 1) == -1) {
                     /* Don't worry, it's just a ping. */
                 }
@@ -1838,14 +2018,8 @@ void replicationCron(void) {
             if (slave->flags & REDIS_PRE_PSYNC) continue;
             if ((server.unixtime - slave->repl_ack_time) > server.repl_timeout)
             {
-                char ip[REDIS_IP_STR_LEN];
-                int port;
-
-                if (anetPeerToString(slave->fd,ip,sizeof(ip),&port) != -1) {
-                    redisLog(REDIS_WARNING,
-                        "Disconnecting timedout slave: %s:%d",
-                        ip, slave->slave_listening_port);
-                }
+                redisLog(REDIS_WARNING, "Disconnecting timedout slave: %s",
+                    replicationGetSlaveName(slave));
                 freeClient(slave);
             }
         }
@@ -1877,6 +2051,48 @@ void replicationCron(void) {
         replicationScriptCacheFlush();
     }
 
+    /* If we are using diskless replication and there are slaves waiting
+     * in WAIT_BGSAVE_START state, check if enough seconds elapsed and
+     * start a BGSAVE.
+     *
+     * This code is also useful to trigger a BGSAVE if the diskless
+     * replication was turned off with CONFIG SET, while there were already
+     * slaves in WAIT_BGSAVE_START state. */
+    if (server.rdb_child_pid == -1 && server.aof_child_pid == -1) {
+        time_t idle, max_idle = 0;
+        int slaves_waiting = 0;
+        listNode *ln;
+        listIter li;
+
+        listRewind(server.slaves,&li);
+        while((ln = listNext(&li))) {
+            redisClient *slave = ln->value;
+            if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_START) {
+                idle = server.unixtime - slave->lastinteraction;
+                if (idle > max_idle) max_idle = idle;
+                slaves_waiting++;
+            }
+        }
+
+        if (slaves_waiting && max_idle > server.repl_diskless_sync_delay) {
+            /* Start a BGSAVE. Usually with socket target, or with disk target
+             * if there was a recent socket -> disk config change. */
+            if (startBgsaveForReplication() == REDIS_OK) {
+                /* It started! We need to change the state of slaves
+                 * from WAIT_BGSAVE_START to WAIT_BGSAVE_END in case
+                 * the current target is disk. Otherwise it was already done
+                 * by rdbSaveToSlavesSockets() which is called by
+                 * startBgsaveForReplication(). */
+                listRewind(server.slaves,&li);
+                while((ln = listNext(&li))) {
+                    redisClient *slave = ln->value;
+                    if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_START)
+                        slave->replstate = REDIS_REPL_WAIT_BGSAVE_END;
+                }
+            }
+        }
+    }
+
     /* Refresh the number of slaves with lag <= min-slaves-max-lag. */
     refreshGoodSlavesCount();
 }

src/rio.c

@@ -55,6 +55,8 @@
 #include "config.h"
 #include "redis.h"
 
+/* ------------------------- Buffer I/O implementation ----------------------- */
+
 /* Returns 1 or 0 for success/failure. */
 static size_t rioBufferWrite(rio *r, const void *buf, size_t len) {
     r->io.buffer.ptr = sdscatlen(r->io.buffer.ptr,(char*)buf,len);
@@ -76,6 +78,33 @@ static off_t rioBufferTell(rio *r) {
     return r->io.buffer.pos;
 }
 
+/* Flushes any buffer to target device if applicable. Returns 1 on success
+ * and 0 on failures. */
+static int rioBufferFlush(rio *r) {
+    REDIS_NOTUSED(r);
+    return 1; /* Nothing to do, our write just appends to the buffer. */
+}
+
+static const rio rioBufferIO = {
+    rioBufferRead,
+    rioBufferWrite,
+    rioBufferTell,
+    rioBufferFlush,
+    NULL,           /* update_checksum */
+    0,              /* current checksum */
+    0,              /* bytes read or written */
+    0,              /* read/write chunk size */
+    { { NULL, 0 } } /* union for io-specific vars */
+};
+
+void rioInitWithBuffer(rio *r, sds s) {
+    *r = rioBufferIO;
+    r->io.buffer.ptr = s;
+    r->io.buffer.pos = 0;
+}
+
+/* --------------------- Stdio file pointer implementation ------------------- */
+
 /* Returns 1 or 0 for success/failure. */
 static size_t rioFileWrite(rio *r, const void *buf, size_t len) {
     size_t retval;
@@ -103,21 +132,17 @@ static off_t rioFileTell(rio *r) {
     return ftello(r->io.file.fp);
 }
 
-static const rio rioBufferIO = {
-    rioBufferRead,
-    rioBufferWrite,
-    rioBufferTell,
-    NULL,           /* update_checksum */
-    0,              /* current checksum */
-    0,              /* bytes read or written */
-    0,              /* read/write chunk size */
-    { { NULL, 0 } } /* union for io-specific vars */
-};
+/* Flushes any buffer to target device if applicable. Returns 1 on success
+ * and 0 on failures. */
+static int rioFileFlush(rio *r) {
+    return (fflush(r->io.file.fp) == 0) ? 1 : 0;
+}
 
 static const rio rioFileIO = {
     rioFileRead,
     rioFileWrite,
     rioFileTell,
+    rioFileFlush,
     NULL,           /* update_checksum */
     0,              /* current checksum */
     0,              /* bytes read or written */
@@ -132,12 +157,133 @@ void rioInitWithFile(rio *r, FILE *fp) {
     r->io.file.autosync = 0;
 }
 
-void rioInitWithBuffer(rio *r, sds s) {
-    *r = rioBufferIO;
-    r->io.buffer.ptr = s;
-    r->io.buffer.pos = 0;
+/* ------------------- File descriptors set implementation ------------------- */
+
+/* Returns 1 or 0 for success/failure.
+ * The function returns success as long as we are able to correctly write
+ * to at least one file descriptor.
+ *
+ * When buf is NULL adn len is 0, the function performs a flush operation
+ * if there is some pending buffer, so this function is also used in order
+ * to implement rioFdsetFlush(). */
+static size_t rioFdsetWrite(rio *r, const void *buf, size_t len) {
+    ssize_t retval;
+    int j;
+    unsigned char *p = (unsigned char*) buf;
+    int doflush = (buf == NULL && len == 0);
+
+    /* To start we always append to our buffer. If it gets larger than
+     * a given size, we actually write to the sockets. */
+    if (len) {
+        r->io.fdset.buf = sdscatlen(r->io.fdset.buf,buf,len);
+        len = 0; /* Prevent entering the while belove if we don't flush. */
+        if (sdslen(r->io.fdset.buf) > REDIS_IOBUF_LEN) doflush = 1;
+    }
+
+    if (doflush) {
+        p = (unsigned char*) r->io.fdset.buf;
+        len = sdslen(r->io.fdset.buf);
+    }
+
+    /* Write in little chunchs so that when there are big writes we
+     * parallelize while the kernel is sending data in background to
+     * the TCP socket. */
+    while(len) {
+        size_t count = len < 1024 ? len : 1024;
+        int broken = 0;
+        for (j = 0; j < r->io.fdset.numfds; j++) {
+            if (r->io.fdset.state[j] != 0) {
+                /* Skip FDs alraedy in error. */
+                broken++;
+                continue;
+            }
+
+            /* Make sure to write 'count' bytes to the socket regardless
+             * of short writes. */
+            size_t nwritten = 0;
+            while(nwritten != count) {
+                retval = write(r->io.fdset.fds[j],p+nwritten,count-nwritten);
+                if (retval <= 0) {
+                    /* With blocking sockets, which is the sole user of this
+                     * rio target, EWOULDBLOCK is returned only because of
+                     * the SO_SNDTIMEO socket option, so we translate the error
+                     * into one more recognizable by the user. */
+                    if (retval == -1 && errno == EWOULDBLOCK) errno = ETIMEDOUT;
+                    break;
+                }
+                nwritten += retval;
+            }
+
+            if (nwritten != count) {
+                /* Mark this FD as broken. */
+                r->io.fdset.state[j] = errno;
+                if (r->io.fdset.state[j] == 0) r->io.fdset.state[j] = EIO;
+            }
+        }
+        if (broken == r->io.fdset.numfds) return 0; /* All the FDs in error. */
+        p += count;
+        len -= count;
+        r->io.fdset.pos += count;
+    }
+
+    if (doflush) sdsclear(r->io.fdset.buf);
+    return 1;
 }
 
+/* Returns 1 or 0 for success/failure. */
+static size_t rioFdsetRead(rio *r, void *buf, size_t len) {
+    REDIS_NOTUSED(r);
+    REDIS_NOTUSED(buf);
+    REDIS_NOTUSED(len);
+    return 0; /* Error, this target does not support reading. */
+}
+
+/* Returns read/write position in file. */
+static off_t rioFdsetTell(rio *r) {
+    return r->io.fdset.pos;
+}
+
+/* Flushes any buffer to target device if applicable. Returns 1 on success
+ * and 0 on failures. */
+static int rioFdsetFlush(rio *r) {
+    /* Our flush is implemented by the write method, that recognizes a
+     * buffer set to NULL with a count of zero as a flush request. */
+    return rioFdsetWrite(r,NULL,0);
+}
+
+static const rio rioFdsetIO = {
+    rioFdsetRead,
+    rioFdsetWrite,
+    rioFdsetTell,
+    rioFdsetFlush,
+    NULL,           /* update_checksum */
+    0,              /* current checksum */
+    0,              /* bytes read or written */
+    0,              /* read/write chunk size */
+    { { NULL, 0 } } /* union for io-specific vars */
+};
+
+void rioInitWithFdset(rio *r, int *fds, int numfds) {
+    int j;
+
+    *r = rioFdsetIO;
+    r->io.fdset.fds = zmalloc(sizeof(int)*numfds);
+    r->io.fdset.state = zmalloc(sizeof(int)*numfds);
+    memcpy(r->io.fdset.fds,fds,sizeof(int)*numfds);
+    for (j = 0; j < numfds; j++) r->io.fdset.state[j] = 0;
+    r->io.fdset.numfds = numfds;
+    r->io.fdset.pos = 0;
+    r->io.fdset.buf = sdsempty();
+}
+
+void rioFreeFdset(rio *r) {
+    zfree(r->io.fdset.fds);
+    zfree(r->io.fdset.state);
+    sdsfree(r->io.fdset.buf);
+}
+
+/* ---------------------------- Generic functions ---------------------------- */
+
 /* This function can be installed both in memory and file streams when checksum
  * computation is needed. */
 void rioGenericUpdateChecksum(rio *r, const void *buf, size_t len) {
@@ -157,7 +303,8 @@ void rioSetAutoSync(rio *r, off_t bytes) {
     r->io.file.autosync = bytes;
 }
 
-/* ------------------------------ Higher level interface ---------------------------
+/* --------------------------- Higher level interface --------------------------
+ *
  * The following higher level functions use lower level rio.c functions to help
  * generating the Redis protocol for the Append Only File. */
 

src/rio.h

@@ -43,6 +43,7 @@ struct _rio {
     size_t (*read)(struct _rio *, void *buf, size_t len);
     size_t (*write)(struct _rio *, const void *buf, size_t len);
     off_t (*tell)(struct _rio *);
+    int (*flush)(struct _rio *);
     /* The update_cksum method if not NULL is used to compute the checksum of
      * all the data that was read or written so far. The method should be
      * designed so that can be called with the current checksum, and the buf
@@ -61,15 +62,25 @@ struct _rio {
 
     /* Backend-specific vars. */
     union {
+        /* In-memory buffer target. */
         struct {
             sds ptr;
             off_t pos;
         } buffer;
+        /* Stdio file pointer target. */
         struct {
             FILE *fp;
             off_t buffered; /* Bytes written since last fsync. */
             off_t autosync; /* fsync after 'autosync' bytes written. */
         } file;
+        /* Multiple FDs target (used to write to N sockets). */
+        struct {
+            int *fds;       /* File descriptors. */
+            int *state;     /* Error state of each fd. 0 (if ok) or errno. */
+            int numfds;
+            off_t pos;
+            sds buf;
+        } fdset;
     } io;
 };
 
@@ -109,8 +120,13 @@ static inline off_t rioTell(rio *r) {
     return r->tell(r);
 }
 
+static inline int rioFlush(rio *r) {
+    return r->flush(r);
+}
+
 void rioInitWithFile(rio *r, FILE *fp);
 void rioInitWithBuffer(rio *r, sds s);
+void rioInitWithFdset(rio *r, int *fds, int numfds);
 
 size_t rioWriteBulkCount(rio *r, char prefix, int count);
 size_t rioWriteBulkString(rio *r, const char *buf, size_t len);

src/scripting.c

@@ -212,7 +212,7 @@ int luaRedisGenericCommand(lua_State *lua, int raise_error) {
     static robj **argv = NULL;
     static int argv_size = 0;
     static robj *cached_objects[LUA_CMD_OBJCACHE_SIZE];
-    static int cached_objects_len[LUA_CMD_OBJCACHE_SIZE];
+    static size_t cached_objects_len[LUA_CMD_OBJCACHE_SIZE];
 
     /* Require at least one argument */
     if (argc == 0) {
@@ -222,9 +222,7 @@ int luaRedisGenericCommand(lua_State *lua, int raise_error) {
     }
 
     /* Build the arguments vector */
-    if (!argv) {
-        argv = zmalloc(sizeof(robj*)*argc);
-    } else if (argv_size < argc) {
+    if (argv_size < argc) {
         argv = zrealloc(argv,sizeof(robj*)*argc);
         argv_size = argc;
     }
@@ -402,6 +400,7 @@ cleanup:
     if (c->argv != argv) {
         zfree(c->argv);
         argv = NULL;
+        argv_size = 0;
     }
 
     if (raise_error) {
@@ -538,6 +537,7 @@ void luaLoadLib(lua_State *lua, const char *libname, lua_CFunction luafunc) {
 LUALIB_API int (luaopen_cjson) (lua_State *L);
 LUALIB_API int (luaopen_struct) (lua_State *L);
 LUALIB_API int (luaopen_cmsgpack) (lua_State *L);
+LUALIB_API int (luaopen_bit) (lua_State *L);
 
 void luaLoadLibraries(lua_State *lua) {
     luaLoadLib(lua, "", luaopen_base);
@@ -548,6 +548,7 @@ void luaLoadLibraries(lua_State *lua) {
     luaLoadLib(lua, "cjson", luaopen_cjson);
     luaLoadLib(lua, "struct", luaopen_struct);
     luaLoadLib(lua, "cmsgpack", luaopen_cmsgpack);
+    luaLoadLib(lua, "bit", luaopen_bit);
 
 #if 0 /* Stuff that we don't load currently, for sandboxing concerns. */
     luaLoadLib(lua, LUA_LOADLIBNAME, luaopen_package);
@@ -718,7 +719,7 @@ void scriptingInit(void) {
         server.lua_client->flags |= REDIS_LUA_CLIENT;
     }
 
-    /* Lua beginners ofter don't use "local", this is likely to introduce
+    /* Lua beginners often don't use "local", this is likely to introduce
      * subtle bugs in their code. To prevent problems we protect accesses
      * to global variables. */
     scriptingEnableGlobalsProtection(lua);
@@ -910,6 +911,9 @@ void evalGenericCommand(redisClient *c, int evalsha) {
     if (numkeys > (c->argc - 3)) {
         addReplyError(c,"Number of keys can't be greater than number of args");
         return;
+    } else if (numkeys < 0) {
+        addReplyError(c,"Number of keys can't be negative");
+        return;
     }
 
     /* We obtain the script SHA1, then check if this function is already

src/sds.c

@@ -43,7 +43,7 @@
  * The string is always null-termined (all the sds strings are, always) so
  * even if you create an sds string with:
  *
- * mystring = sdsnewlen("abc",3");
+ * mystring = sdsnewlen("abc",3);
  *
  * You can print the string with printf() as there is an implicit \0 at the
  * end of the string. However the string is binary safe and can contain
@@ -109,7 +109,7 @@ void sdsupdatelen(sds s) {
     sh->len = reallen;
 }
 
-/* Modify an sds string on-place to make it empty (zero length).
+/* Modify an sds string in-place to make it empty (zero length).
  * However all the existing buffer is not discarded but set as free space
  * so that next append operations will not require allocations up to the
  * number of bytes previously available. */
@@ -200,10 +200,12 @@ size_t sdsAllocSize(sds s) {
 void sdsIncrLen(sds s, int incr) {
     struct sdshdr *sh = (void*) (s-(sizeof(struct sdshdr)));
 
-    assert(sh->free >= incr);
+    if (incr >= 0)
+        assert(sh->free >= (unsigned int)incr);
+    else
+        assert(sh->len >= (unsigned int)(-incr));
     sh->len += incr;
     sh->free -= incr;
-    assert(sh->free >= 0);
     s[sh->len] = '\0';
 }
 
@@ -388,6 +390,7 @@ sds sdscatvprintf(sds s, const char *fmt, va_list ap) {
         buf[buflen-2] = '\0';
         va_copy(cpy,ap);
         vsnprintf(buf, buflen, fmt, cpy);
+        va_end(ap);
         if (buf[buflen-2] != '\0') {
             if (buf != staticbuf) zfree(buf);
             buflen *= 2;
@@ -457,7 +460,7 @@ sds sdscatfmt(sds s, char const *fmt, ...) {
     i = initlen; /* Position of the next byte to write to dest str. */
     while(*f) {
         char next, *str;
-        int l;
+        unsigned int l;
         long long num;
         unsigned long long unum;
 

src/sds.h

@@ -39,8 +39,8 @@
 typedef char *sds;
 
 struct sdshdr {
-    int len;
-    int free;
+    unsigned int len;
+    unsigned int free;
     char buf[];
 };
 

src/sentinel.c

@@ -159,7 +159,7 @@ typedef struct sentinelRedisInstance {
     /* Master specific. */
     dict *sentinels;    /* Other sentinels monitoring the same master. */
     dict *slaves;       /* Slaves for this master instance. */
-    int quorum;         /* Number of sentinels that need to agree on failure. */
+    unsigned int quorum;/* Number of sentinels that need to agree on failure. */
     int parallel_syncs; /* How many slaves to reconfigure at same time. */
     char *auth_pass;    /* Password to use for AUTH against master & slaves. */
 
@@ -203,6 +203,10 @@ struct sentinelState {
     mstime_t tilt_start_time;   /* When TITL started. */
     mstime_t previous_time;     /* Last time we ran the time handler. */
     list *scripts_queue;    /* Queue of user scripts to execute. */
+    char *announce_ip;      /* IP addr that is gossiped to other sentinels if
+                               not NULL. */
+    int announce_port;      /* Port that is gossiped to other sentinels if
+                               non zero. */
 } sentinel;
 
 /* A script execution job. */
@@ -345,6 +349,7 @@ int dictSdsKeyCompare(void *privdata, const void *key1, const void *key2);
 void releaseSentinelRedisInstance(sentinelRedisInstance *ri);
 
 void dictInstancesValDestructor (void *privdata, void *obj) {
+    REDIS_NOTUSED(privdata);
     releaseSentinelRedisInstance(obj);
 }
 
@@ -403,7 +408,7 @@ void initSentinelConfig(void) {
 
 /* Perform the Sentinel mode initialization. */
 void initSentinel(void) {
-    int j;
+    unsigned int j;
 
     /* Remove usual Redis commands from the command table, then just add
      * the SENTINEL command. */
@@ -424,6 +429,8 @@ void initSentinel(void) {
     sentinel.previous_time = mstime();
     sentinel.running_scripts = 0;
     sentinel.scripts_queue = listCreate();
+    sentinel.announce_ip = NULL;
+    sentinel.announce_port = 0;
 }
 
 /* This function gets called when the server is in Sentinel mode, started,
@@ -455,19 +462,19 @@ void sentinelIsRunning(void) {
  *  EINVAL: Invalid port number.
  */
 sentinelAddr *createSentinelAddr(char *hostname, int port) {
-    char buf[32];
+    char ip[REDIS_IP_STR_LEN];
     sentinelAddr *sa;
 
     if (port <= 0 || port > 65535) {
         errno = EINVAL;
         return NULL;
     }
-    if (anetResolve(NULL,hostname,buf,sizeof(buf)) == ANET_ERR) {
+    if (anetResolve(NULL,hostname,ip,sizeof(ip)) == ANET_ERR) {
         errno = ENOENT;
         return NULL;
     }
     sa = zmalloc(sizeof(*sa));
-    sa->ip = sdsnew(buf);
+    sa->ip = sdsnew(ip);
     sa->port = port;
     return sa;
 }
@@ -1424,6 +1431,13 @@ char *sentinelHandleConfiguration(char **argv, int argc) {
             return "Wrong hostname or port for sentinel.";
         }
         if (argc == 5) si->runid = sdsnew(argv[4]);
+    } else if (!strcasecmp(argv[0],"announce-ip") && argc == 2) {
+        /* announce-ip <ip-address> */
+        if (strlen(argv[1]))
+            sentinel.announce_ip = sdsnew(argv[1]);
+    } else if (!strcasecmp(argv[0],"announce-port") && argc == 2) {
+        /* announce-port <port> */
+        sentinel.announce_port = atoi(argv[1]);
     } else {
         return "Unrecognized sentinel configuration statement.";
     }
@@ -1555,6 +1569,20 @@ void rewriteConfigSentinelOption(struct rewriteConfigState *state) {
         "sentinel current-epoch %llu", (unsigned long long) sentinel.current_epoch);
     rewriteConfigRewriteLine(state,"sentinel",line,1);
 
+    /* sentinel announce-ip. */
+    if (sentinel.announce_ip) {
+        line = sdsnew("sentinel announce-ip ");
+        line = sdscatrepr(line, sentinel.announce_ip, sdslen(sentinel.announce_ip));
+        rewriteConfigRewriteLine(state,"sentinel",line,1);
+    }
+
+    /* sentinel announce-port. */
+    if (sentinel.announce_port) {
+        line = sdscatprintf(sdsempty(),"sentinel announce-port %d",
+                            sentinel.announce_port);
+        rewriteConfigRewriteLine(state,"sentinel",line,1);
+    }
+
     dictReleaseIterator(di);
 }
 
@@ -1634,6 +1662,7 @@ void sentinelLinkEstablishedCallback(const redisAsyncContext *c, int status) {
 }
 
 void sentinelDisconnectCallback(const redisAsyncContext *c, int status) {
+    REDIS_NOTUSED(status);
     sentinelDisconnectInstanceFromContext(c);
 }
 
@@ -1998,6 +2027,7 @@ void sentinelRefreshInstanceInfo(sentinelRedisInstance *ri, const char *info) {
 void sentinelInfoReplyCallback(redisAsyncContext *c, void *reply, void *privdata) {
     sentinelRedisInstance *ri = c->data;
     redisReply *r;
+    REDIS_NOTUSED(privdata);
 
     if (ri) ri->pending_commands--;
     if (!reply || !ri) return;
@@ -2012,6 +2042,8 @@ void sentinelInfoReplyCallback(redisAsyncContext *c, void *reply, void *privdata
  * value of the command but its effects directly. */
 void sentinelDiscardReplyCallback(redisAsyncContext *c, void *reply, void *privdata) {
     sentinelRedisInstance *ri = c->data;
+    REDIS_NOTUSED(reply);
+    REDIS_NOTUSED(privdata);
 
     if (ri) ri->pending_commands--;
 }
@@ -2019,6 +2051,7 @@ void sentinelDiscardReplyCallback(redisAsyncContext *c, void *reply, void *privd
 void sentinelPingReplyCallback(redisAsyncContext *c, void *reply, void *privdata) {
     sentinelRedisInstance *ri = c->data;
     redisReply *r;
+    REDIS_NOTUSED(privdata);
 
     if (ri) ri->pending_commands--;
     if (!reply || !ri) return;
@@ -2057,6 +2090,7 @@ void sentinelPingReplyCallback(redisAsyncContext *c, void *reply, void *privdata
 void sentinelPublishReplyCallback(redisAsyncContext *c, void *reply, void *privdata) {
     sentinelRedisInstance *ri = c->data;
     redisReply *r;
+    REDIS_NOTUSED(privdata);
 
     if (ri) ri->pending_commands--;
     if (!reply || !ri) return;
@@ -2072,7 +2106,7 @@ void sentinelPublishReplyCallback(redisAsyncContext *c, void *reply, void *privd
  * or sent directly to this sentinel via the (fake) PUBLISH command of Sentinel.
  *
  * If the master name specified in the message is not known, the message is
- * discareded. */
+ * discarded. */
 void sentinelProcessHelloMessage(char *hello, int hello_len) {
     /* Format is composed of 8 tokens:
      * 0=ip,1=port,2=runid,3=current_epoch,4=master_name,
@@ -2166,6 +2200,7 @@ cleanup:
 void sentinelReceiveHelloMessages(redisAsyncContext *c, void *reply, void *privdata) {
     sentinelRedisInstance *ri = c->data;
     redisReply *r;
+    REDIS_NOTUSED(privdata);
 
     if (!reply || !ri) return;
     r = reply;
@@ -2205,18 +2240,30 @@ int sentinelSendHello(sentinelRedisInstance *ri) {
     char ip[REDIS_IP_STR_LEN];
     char payload[REDIS_IP_STR_LEN+1024];
     int retval;
+    char *announce_ip;
+    int announce_port;
     sentinelRedisInstance *master = (ri->flags & SRI_MASTER) ? ri : ri->master;
     sentinelAddr *master_addr = sentinelGetCurrentMasterAddress(master);
 
-    /* Try to obtain our own IP address. */
-    if (anetSockName(ri->cc->c.fd,ip,sizeof(ip),NULL) == -1) return REDIS_ERR;
     if (ri->flags & SRI_DISCONNECTED) return REDIS_ERR;
 
+    /* Use the specified announce address if specified, otherwise try to
+     * obtain our own IP address. */
+    if (sentinel.announce_ip) {
+        announce_ip = sentinel.announce_ip;
+    } else {
+        if (anetSockName(ri->cc->c.fd,ip,sizeof(ip),NULL) == -1)
+            return REDIS_ERR;
+        announce_ip = ip;
+    }
+    announce_port = sentinel.announce_port ?
+                    sentinel.announce_port : server.port;
+
     /* Format and send the Hello message. */
     snprintf(payload,sizeof(payload),
         "%s,%d,%s,%llu," /* Info about this sentinel. */
         "%s,%s,%d,%llu", /* Info about current master. */
-        ip, server.port, server.runid,
+        announce_ip, announce_port, server.runid,
         (unsigned long long) sentinel.current_epoch,
         /* --- */
         master->name,master_addr->ip,master_addr->port,
@@ -2559,7 +2606,7 @@ sentinelRedisInstance *sentinelGetMasterByNameOrReplyError(redisClient *c,
 {
     sentinelRedisInstance *ri;
 
-    ri = dictFetchValue(sentinel.masters,c->argv[2]->ptr);
+    ri = dictFetchValue(sentinel.masters,name->ptr);
     if (!ri) {
         addReplyError(c,"No such master with that name");
         return NULL;
@@ -2682,7 +2729,7 @@ void sentinelCommand(redisClient *c) {
         /* SENTINEL MONITOR <name> <ip> <port> <quorum> */
         sentinelRedisInstance *ri;
         long quorum, port;
-        char buf[32];
+        char ip[REDIS_IP_STR_LEN];
 
         if (c->argc != 6) goto numargserr;
         if (getLongFromObjectOrReply(c,c->argv[5],&quorum,"Invalid quorum")
@@ -2692,7 +2739,7 @@ void sentinelCommand(redisClient *c) {
         /* Make sure the IP field is actually a valid IP before passing it
          * to createSentinelRedisInstance(), otherwise we may trigger a
          * DNS lookup at runtime. */
-        if (anetResolveIP(NULL,c->argv[3]->ptr,buf,sizeof(buf)) == ANET_ERR) {
+        if (anetResolveIP(NULL,c->argv[3]->ptr,ip,sizeof(ip)) == ANET_ERR) {
             addReplyError(c,"Invalid IP address specified");
             return;
         }
@@ -2997,7 +3044,7 @@ void sentinelCheckSubjectivelyDown(sentinelRedisInstance *ri) {
 void sentinelCheckObjectivelyDown(sentinelRedisInstance *master) {
     dictIterator *di;
     dictEntry *de;
-    int quorum = 0, odown = 0;
+    unsigned int quorum = 0, odown = 0;
 
     if (master->flags & SRI_S_DOWN) {
         /* Is down for enough sentinels? */
@@ -3034,6 +3081,7 @@ void sentinelCheckObjectivelyDown(sentinelRedisInstance *master) {
 void sentinelReceiveIsMasterDownReply(redisAsyncContext *c, void *reply, void *privdata) {
     sentinelRedisInstance *ri = c->data;
     redisReply *r;
+    REDIS_NOTUSED(privdata);
 
     if (ri) ri->pending_commands--;
     if (!reply || !ri) return;
@@ -3057,7 +3105,7 @@ void sentinelReceiveIsMasterDownReply(redisAsyncContext *c, void *reply, void *p
             /* If the runid in the reply is not "*" the Sentinel actually
              * replied with a vote. */
             sdsfree(ri->leader);
-            if (ri->leader_epoch != r->element[2]->integer)
+            if ((long long)ri->leader_epoch != r->element[2]->integer)
                 redisLog(REDIS_WARNING,
                     "%s voted for %s %llu", ri->name,
                     r->element[1]->str,
@@ -3176,9 +3224,9 @@ int sentinelLeaderIncr(dict *counters, char *runid) {
 /* Scan all the Sentinels attached to this master to check if there
  * is a leader for the specified epoch.
  *
- * To be a leader for a given epoch, we should have the majorify of
- * the Sentinels we know that reported the same instance as
- * leader for the same epoch. */
+ * To be a leader for a given epoch, we should have the majority of
+ * the Sentinels we know (ever seen since the last SENTINEL RESET) that
+ * reported the same instance as leader for the same epoch. */
 char *sentinelGetLeader(sentinelRedisInstance *master, uint64_t epoch) {
     dict *counters;
     dictIterator *di;
@@ -3192,13 +3240,14 @@ char *sentinelGetLeader(sentinelRedisInstance *master, uint64_t epoch) {
     redisAssert(master->flags & (SRI_O_DOWN|SRI_FAILOVER_IN_PROGRESS));
     counters = dictCreate(&leaderVotesDictType,NULL);
 
+    voters = dictSize(master->sentinels)+1; /* All the other sentinels and me. */
+
     /* Count other sentinels votes */
     di = dictGetIterator(master->sentinels);
     while((de = dictNext(di)) != NULL) {
         sentinelRedisInstance *ri = dictGetVal(de);
         if (ri->leader != NULL && ri->leader_epoch == sentinel.current_epoch)
             sentinelLeaderIncr(counters,ri->leader);
-        voters++;
     }
     dictReleaseIterator(di);
 
@@ -3232,7 +3281,6 @@ char *sentinelGetLeader(sentinelRedisInstance *master, uint64_t epoch) {
             winner = myvote;
         }
     }
-    voters++; /* Anyway, count me as one of the voters. */
 
     voters_quorum = voters/2+1;
     if (winner && (max_votes < voters_quorum || max_votes < master->quorum))

src/syncio.c

@@ -139,6 +139,7 @@ ssize_t syncReadLine(int fd, char *ptr, ssize_t size, long long timeout) {
             *ptr = '\0';
             nread++;
         }
+        size--;
     }
     return nread;
 }

src/t_hash.c

@@ -565,7 +565,7 @@ void hincrbyfloatCommand(redisClient *c) {
     }
 
     value += incr;
-    new = createStringObjectFromLongDouble(value);
+    new = createStringObjectFromLongDouble(value,1);
     hashTypeTryObjectEncoding(o,&c->argv[2],NULL);
     hashTypeSet(o,c->argv[2],new);
     addReplyBulk(c,new);

src/t_string.c

@@ -69,7 +69,7 @@ void setGenericCommand(redisClient *c, int flags, robj *key, robj *val, robj *ex
         if (getLongLongFromObjectOrReply(c, expire, &milliseconds, NULL) != REDIS_OK)
             return;
         if (milliseconds <= 0) {
-            addReplyError(c,"invalid expire time in SETEX");
+            addReplyErrorFormat(c,"invalid expire time in %s",c->cmd->name);
             return;
         }
         if (unit == UNIT_SECONDS) milliseconds *= 1000;
@@ -231,13 +231,13 @@ void setrangeCommand(redisClient *c) {
 
 void getrangeCommand(redisClient *c) {
     robj *o;
-    long start, end;
+    long long start, end;
     char *str, llbuf[32];
     size_t strlen;
 
-    if (getLongFromObjectOrReply(c,c->argv[2],&start,NULL) != REDIS_OK)
+    if (getLongLongFromObjectOrReply(c,c->argv[2],&start,NULL) != REDIS_OK)
         return;
-    if (getLongFromObjectOrReply(c,c->argv[3],&end,NULL) != REDIS_OK)
+    if (getLongLongFromObjectOrReply(c,c->argv[3],&end,NULL) != REDIS_OK)
         return;
     if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.emptybulk)) == NULL ||
         checkType(c,o,REDIS_STRING)) return;
@@ -255,11 +255,11 @@ void getrangeCommand(redisClient *c) {
     if (end < 0) end = strlen+end;
     if (start < 0) start = 0;
     if (end < 0) end = 0;
-    if ((unsigned)end >= strlen) end = strlen-1;
+    if ((unsigned long long)end >= strlen) end = strlen-1;
 
     /* Precondition: end >= 0 && end < strlen, so the only condition where
      * nothing can be returned is: start > end. */
-    if (start > end) {
+    if (start > end || strlen == 0) {
         addReply(c,shared.emptybulk);
     } else {
         addReplyBulkCBuffer(c,(char*)str+start,end-start+1);
@@ -337,11 +337,21 @@ void incrDecrCommand(redisClient *c, long long incr) {
         return;
     }
     value += incr;
-    new = createStringObjectFromLongLong(value);
-    if (o)
-        dbOverwrite(c->db,c->argv[1],new);
-    else
-        dbAdd(c->db,c->argv[1],new);
+
+    if (o && o->refcount == 1 && o->encoding == REDIS_ENCODING_INT &&
+        (value < 0 || value >= REDIS_SHARED_INTEGERS) &&
+        value >= LONG_MIN && value <= LONG_MAX)
+    {
+        new = o;
+        o->ptr = (void*)((long)value);
+    } else {
+        new = createStringObjectFromLongLong(value);
+        if (o) {
+            dbOverwrite(c->db,c->argv[1],new);
+        } else {
+            dbAdd(c->db,c->argv[1],new);
+        }
+    }
     signalModifiedKey(c->db,c->argv[1]);
     notifyKeyspaceEvent(REDIS_NOTIFY_STRING,"incrby",c->argv[1],c->db->id);
     server.dirty++;
@@ -387,7 +397,7 @@ void incrbyfloatCommand(redisClient *c) {
         addReplyError(c,"increment would produce NaN or Infinity");
         return;
     }
-    new = createStringObjectFromLongDouble(value);
+    new = createStringObjectFromLongDouble(value,1);
     if (o)
         dbOverwrite(c->db,c->argv[1],new);
     else

src/t_zset.c

@@ -205,8 +205,6 @@ int zslDelete(zskiplist *zsl, double score, robj *obj) {
         zslDeleteNode(zsl, x, update);
         zslFreeNode(x);
         return 1;
-    } else {
-        return 0; /* not found */
     }
     return 0; /* not found */
 }
@@ -1981,51 +1979,77 @@ void zunionInterGenericCommand(redisClient *c, robj *dstkey, int op) {
             zuiClearIterator(&src[0]);
         }
     } else if (op == REDIS_OP_UNION) {
+        dict *accumulator = dictCreate(&setDictType,NULL);
+        dictIterator *di;
+        dictEntry *de;
+        double score;
+
+        if (setnum) {
+            /* Our union is at least as large as the largest set.
+             * Resize the dictionary ASAP to avoid useless rehashing. */
+            dictExpand(accumulator,zuiLength(&src[setnum-1]));
+        }
+
+        /* Step 1: Create a dictionary of elements -> aggregated-scores
+         * by iterating one sorted set after the other. */
         for (i = 0; i < setnum; i++) {
-            if (zuiLength(&src[i]) == 0)
-                continue;
+            if (zuiLength(&src[i]) == 0) continue;
 
             zuiInitIterator(&src[i]);
             while (zuiNext(&src[i],&zval)) {
-                double score, value;
-
-                /* Skip an element that when already processed */
-                if (dictFind(dstzset->dict,zuiObjectFromValue(&zval)) != NULL)
-                    continue;
-
-                /* Initialize score */
+                /* Initialize value */
                 score = src[i].weight * zval.score;
                 if (isnan(score)) score = 0;
 
-                /* We need to check only next sets to see if this element
-                 * exists, since we process every element just one time so
-                 * it can't exist in a previous set (otherwise it would be
-                 * already processed). */
-                for (j = (i+1); j < setnum; j++) {
-                    /* It is not safe to access the zset we are
-                     * iterating, so explicitly check for equal object. */
-                    if(src[j].subject == src[i].subject) {
-                        value = zval.score*src[j].weight;
-                        zunionInterAggregate(&score,value,aggregate);
-                    } else if (zuiFind(&src[j],&zval,&value)) {
-                        value *= src[j].weight;
-                        zunionInterAggregate(&score,value,aggregate);
+                /* Search for this element in the accumulating dictionary. */
+                de = dictFind(accumulator,zuiObjectFromValue(&zval));
+                /* If we don't have it, we need to create a new entry. */
+                if (de == NULL) {
+                    tmp = zuiObjectFromValue(&zval);
+                    /* Remember the longest single element encountered,
+                     * to understand if it's possible to convert to ziplist
+                     * at the end. */
+                    if (sdsEncodedObject(tmp)) {
+                        if (sdslen(tmp->ptr) > maxelelen)
+                            maxelelen = sdslen(tmp->ptr);
                     }
-                }
-
-                tmp = zuiObjectFromValue(&zval);
-                znode = zslInsert(dstzset->zsl,score,tmp);
-                incrRefCount(zval.ele); /* added to skiplist */
-                dictAdd(dstzset->dict,tmp,&znode->score);
-                incrRefCount(zval.ele); /* added to dictionary */
-
-                if (sdsEncodedObject(tmp)) {
-                    if (sdslen(tmp->ptr) > maxelelen)
-                        maxelelen = sdslen(tmp->ptr);
+                    /* Add the element with its initial score. */
+                    de = dictAddRaw(accumulator,tmp);
+                    incrRefCount(tmp);
+                    dictSetDoubleVal(de,score);
+                } else {
+                    /* Update the score with the score of the new instance
+                     * of the element found in the current sorted set.
+                     *
+                     * Here we access directly the dictEntry double
+                     * value inside the union as it is a big speedup
+                     * compared to using the getDouble/setDouble API. */
+                    zunionInterAggregate(&de->v.d,score,aggregate);
                 }
             }
             zuiClearIterator(&src[i]);
         }
+
+        /* Step 2: convert the dictionary into the final sorted set. */
+        di = dictGetIterator(accumulator);
+
+        /* We now are aware of the final size of the resulting sorted set,
+         * let's resize the dictionary embedded inside the sorted set to the
+         * right size, in order to save rehashing time. */
+        dictExpand(dstzset->dict,dictSize(accumulator));
+
+        while((de = dictNext(di)) != NULL) {
+            robj *ele = dictGetKey(de);
+            score = dictGetDoubleVal(de);
+            znode = zslInsert(dstzset->zsl,score,ele);
+            incrRefCount(ele); /* added to skiplist */
+            dictAdd(dstzset->dict,ele,&znode->score);
+            incrRefCount(ele); /* added to dictionary */
+        }
+        dictReleaseIterator(di);
+
+        /* We can free the accumulator dictionary now. */
+        dictRelease(accumulator);
     } else {
         redisPanic("Unknown operator");
     }

src/util.c

@@ -37,6 +37,7 @@
 #include <unistd.h>
 #include <sys/time.h>
 #include <float.h>
+#include <stdint.h>
 
 #include "util.h"
 
@@ -214,28 +215,91 @@ long long memtoll(const char *p, int *err) {
     return val*mul;
 }
 
-/* Convert a long long into a string. Returns the number of
- * characters needed to represent the number, that can be shorter if passed
- * buffer length is not enough to store the whole number. */
-int ll2string(char *s, size_t len, long long value) {
-    char buf[32], *p;
-    unsigned long long v;
-    size_t l;
+/* Return the number of digits of 'v' when converted to string in radix 10.
+ * See ll2string() for more information. */
+uint32_t digits10(uint64_t v) {
+    if (v < 10) return 1;
+    if (v < 100) return 2;
+    if (v < 1000) return 3;
+    if (v < 1000000000000UL) {
+        if (v < 100000000UL) {
+            if (v < 1000000) {
+                if (v < 10000) return 4;
+                return 5 + (v >= 100000);
+            }
+            return 7 + (v >= 10000000UL);
+        }
+        if (v < 10000000000UL) {
+            return 9 + (v >= 1000000000UL);
+        }
+        return 11 + (v >= 100000000000UL);
+    }
+    return 12 + digits10(v / 1000000000000UL);
+}
 
-    if (len == 0) return 0;
-    v = (value < 0) ? -value : value;
-    p = buf+31; /* point to the last character */
-    do {
-        *p-- = '0'+(v%10);
-        v /= 10;
-    } while(v);
-    if (value < 0) *p-- = '-';
-    p++;
-    l = 32-(p-buf);
-    if (l+1 > len) l = len-1; /* Make sure it fits, including the nul term */
-    memcpy(s,p,l);
-    s[l] = '\0';
-    return l;
+/* Convert a long long into a string. Returns the number of
+ * characters needed to represent the number.
+ * If the buffer is not big enough to store the string, 0 is returned.
+ *
+ * Based on the following article (that apparently does not provide a
+ * novel approach but only publicizes an already used technique):
+ *
+ * https://www.facebook.com/notes/facebook-engineering/three-optimization-tips-for-c/10151361643253920
+ *
+ * Modified in order to handle signed integers since the original code was
+ * designed for unsigned integers. */
+int ll2string(char* dst, size_t dstlen, long long svalue) {
+    static const char digits[201] =
+        "0001020304050607080910111213141516171819"
+        "2021222324252627282930313233343536373839"
+        "4041424344454647484950515253545556575859"
+        "6061626364656667686970717273747576777879"
+        "8081828384858687888990919293949596979899";
+    int negative;
+    unsigned long long value;
+
+    /* The main loop works with 64bit unsigned integers for simplicity, so
+     * we convert the number here and remember if it is negative. */
+    if (svalue < 0) {
+        if (svalue != LLONG_MIN) {
+            value = -svalue;
+        } else {
+            value = ((unsigned long long) LLONG_MAX)+1;
+        }
+        negative = 1;
+    } else {
+        value = svalue;
+        negative = 0;
+    }
+
+    /* Check length. */
+    uint32_t const length = digits10(value)+negative;
+    if (length >= dstlen) return 0;
+
+    /* Null term. */
+    uint32_t next = length;
+    dst[next] = '\0';
+    next--;
+    while (value >= 100) {
+        int const i = (value % 100) * 2;
+        value /= 100;
+        dst[next] = digits[i + 1];
+        dst[next - 1] = digits[i];
+        next -= 2;
+    }
+
+    /* Handle last 1-2 digits. */
+    if (value < 10) {
+        dst[next] = '0' + (uint32_t) value;
+    } else {
+        int i = (uint32_t) value * 2;
+        dst[next] = digits[i + 1];
+        dst[next - 1] = digits[i];
+    }
+
+    /* Add sign. */
+    if (negative) dst[0] = '-';
+    return length;
 }
 
 /* Convert a string into a long long. Returns 1 if the string could be parsed
@@ -321,7 +385,7 @@ int string2l(const char *s, size_t slen, long *lval) {
 }
 
 /* Convert a double to a string representation. Returns the number of bytes
- * required. The representation should always be parsable by stdtod(3). */
+ * required. The representation should always be parsable by strtod(3). */
 int d2string(char *buf, size_t len, double value) {
     if (isnan(value)) {
         len = snprintf(buf,len,"nan");

src/ziplist.c

@@ -183,7 +183,7 @@ static unsigned int zipIntSize(unsigned char encoding) {
     return 0;
 }
 
-/* Encode the length 'l' writing it in 'p'. If p is NULL it just returns
+/* Encode the length 'rawlen' writing it in 'p'. If p is NULL it just returns
  * the amount of bytes required to encode such a length. */
 static unsigned int zipEncodeLength(unsigned char *p, unsigned char encoding, unsigned int rawlen) {
     unsigned char len = 1, buf[5];
@@ -576,19 +576,19 @@ static unsigned char *__ziplistDelete(unsigned char *zl, unsigned char *p, unsig
 
 /* Insert item at "p". */
 static unsigned char *__ziplistInsert(unsigned char *zl, unsigned char *p, unsigned char *s, unsigned int slen) {
-    size_t curlen = intrev32ifbe(ZIPLIST_BYTES(zl)), reqlen, prevlen = 0;
+    size_t curlen = intrev32ifbe(ZIPLIST_BYTES(zl)), reqlen;
+    unsigned int prevlensize, prevlen = 0;
     size_t offset;
     int nextdiff = 0;
     unsigned char encoding = 0;
     long long value = 123456789; /* initialized to avoid warning. Using a value
                                     that is easy to see if for some reason
                                     we use it uninitialized. */
-    zlentry entry, tail;
+    zlentry tail;
 
     /* Find out prevlen for the entry that is inserted. */
     if (p[0] != ZIP_END) {
-        entry = zipEntry(p);
-        prevlen = entry.prevrawlen;
+        ZIP_DECODE_PREVLEN(p, prevlensize, prevlen);
     } else {
         unsigned char *ptail = ZIPLIST_ENTRY_TAIL(zl);
         if (ptail[0] != ZIP_END) {
@@ -676,15 +676,15 @@ unsigned char *ziplistPush(unsigned char *zl, unsigned char *s, unsigned int sle
  * doesn't contain an element at the provided index, NULL is returned. */
 unsigned char *ziplistIndex(unsigned char *zl, int index) {
     unsigned char *p;
-    zlentry entry;
+    unsigned int prevlensize, prevlen = 0;
     if (index < 0) {
         index = (-index)-1;
         p = ZIPLIST_ENTRY_TAIL(zl);
         if (p[0] != ZIP_END) {
-            entry = zipEntry(p);
-            while (entry.prevrawlen > 0 && index--) {
-                p -= entry.prevrawlen;
-                entry = zipEntry(p);
+            ZIP_DECODE_PREVLEN(p, prevlensize, prevlen);
+            while (prevlen > 0 && index--) {
+                p -= prevlen;
+                ZIP_DECODE_PREVLEN(p, prevlensize, prevlen);
             }
         }
     } else {
@@ -722,7 +722,7 @@ unsigned char *ziplistNext(unsigned char *zl, unsigned char *p) {
 
 /* Return pointer to previous entry in ziplist. */
 unsigned char *ziplistPrev(unsigned char *zl, unsigned char *p) {
-    zlentry entry;
+    unsigned int prevlensize, prevlen = 0;
 
     /* Iterating backwards from ZIP_END should return the tail. When "p" is
      * equal to the first element of the list, we're already at the head,
@@ -733,14 +733,14 @@ unsigned char *ziplistPrev(unsigned char *zl, unsigned char *p) {
     } else if (p == ZIPLIST_ENTRY_HEAD(zl)) {
         return NULL;
     } else {
-        entry = zipEntry(p);
-        assert(entry.prevrawlen > 0);
-        return p-entry.prevrawlen;
+        ZIP_DECODE_PREVLEN(p, prevlensize, prevlen);
+        assert(prevlen > 0);
+        return p-prevlen;
     }
 }
 
-/* Get entry pointed to by 'p' and store in either 'e' or 'v' depending
- * on the encoding of the entry. 'e' is always set to NULL to be able
+/* Get entry pointed to by 'p' and store in either '*sstr' or 'sval' depending
+ * on the encoding of the entry. '*sstr' is always set to NULL to be able
  * to find out whether the string pointer or the integer value was set.
  * Return 0 if 'p' points to the end of the ziplist, 1 otherwise. */
 unsigned int ziplistGet(unsigned char *p, unsigned char **sstr, unsigned int *slen, long long *sval) {
@@ -788,7 +788,8 @@ unsigned char *ziplistDeleteRange(unsigned char *zl, unsigned int index, unsigne
     return (p == NULL) ? zl : __ziplistDelete(zl,p,num);
 }
 
-/* Compare entry pointer to by 'p' with 'entry'. Return 1 if equal. */
+/* Compare entry pointer to by 'p' with 'sstr' of length 'slen'. */
+/* Return 1 if equal. */
 unsigned int ziplistCompare(unsigned char *p, unsigned char *sstr, unsigned int slen) {
     zlentry entry;
     unsigned char sencoding;

src/zmalloc.c

@@ -328,27 +328,39 @@ float zmalloc_get_fragmentation_ratio(size_t rss) {
     return (float)rss/zmalloc_used_memory();
 }
 
+/* Get the sum of the specified field (converted form kb to bytes) in
+ * /proc/self/smaps. The field must be specified with trailing ":" as it
+ * apperas in the smaps output.
+ *
+ * Example: zmalloc_get_smap_bytes_by_field("Rss:");
+ */
 #if defined(HAVE_PROC_SMAPS)
-size_t zmalloc_get_private_dirty(void) {
+size_t zmalloc_get_smap_bytes_by_field(char *field) {
     char line[1024];
-    size_t pd = 0;
+    size_t bytes = 0;
     FILE *fp = fopen("/proc/self/smaps","r");
+    int flen = strlen(field);
 
     if (!fp) return 0;
     while(fgets(line,sizeof(line),fp) != NULL) {
-        if (strncmp(line,"Private_Dirty:",14) == 0) {
+        if (strncmp(line,field,flen) == 0) {
             char *p = strchr(line,'k');
             if (p) {
                 *p = '\0';
-                pd += strtol(line+14,NULL,10) * 1024;
+                bytes += strtol(line+flen,NULL,10) * 1024;
             }
         }
     }
     fclose(fp);
-    return pd;
+    return bytes;
 }
 #else
-size_t zmalloc_get_private_dirty(void) {
+size_t zmalloc_get_smap_bytes_by_field(char *field) {
+    ((void) field);
     return 0;
 }
 #endif
+
+size_t zmalloc_get_private_dirty(void) {
+    return zmalloc_get_smap_bytes_by_field("Private_Dirty:");
+}

src/zmalloc.h

@@ -76,6 +76,7 @@ void zmalloc_set_oom_handler(void (*oom_handler)(size_t));
 float zmalloc_get_fragmentation_ratio(size_t rss);
 size_t zmalloc_get_rss(void);
 size_t zmalloc_get_private_dirty(void);
+size_t zmalloc_get_smap_bytes_by_field(char *field);
 void zlibc_free(void *ptr);
 
 #ifndef HAVE_MALLOC_SIZE

tests/assets/default.conf

@@ -9,6 +9,7 @@ bind 127.0.0.1
 loglevel verbose
 logfile ''
 databases 16
+latency-monitor-threshold 1
 
 save 900 1
 save 300 10

tests/cluster/cluster.tcl

@@ -1,7 +1,7 @@
 # Cluster-specific test functions.
 #
 # Copyright (C) 2014 Salvatore Sanfilippo antirez@gmail.com
-# This softare is released under the BSD License. See the COPYING file for
+# This software is released under the BSD License. See the COPYING file for
 # more information.
 
 # Returns a parsed CLUSTER NODES output as a list of dictionaries.

tests/cluster/run.tcl

@@ -1,5 +1,5 @@
 # Cluster test suite. Copyright (C) 2014 Salvatore Sanfilippo antirez@gmail.com
-# This softare is released under the BSD License. See the COPYING file for
+# This software is released under the BSD License. See the COPYING file for
 # more information.
 
 cd tests/cluster

tests/cluster/tests/03-failover-loop.tcl

@@ -49,6 +49,18 @@ while {[incr iterations -1]} {
         set slave_config_epoch [CI $slave cluster_my_epoch]
     }
 
+    test "Cluster is writable before failover" {
+        for {set i 0} {$i < 100} {incr i} {
+            catch {$cluster set $key:$i $val:$i} err
+            assert {$err eq {OK}}
+        }
+        # Wait for the write to propagate to the slave if we
+        # are going to kill a master.
+        if {$role eq {master}} {
+            R $tokill wait 1 20000
+        }
+    }
+
     test "Killing node #$tokill" {
         kill_instance redis $tokill
     }
@@ -67,9 +79,9 @@ while {[incr iterations -1]} {
         assert_cluster_state ok
     }
 
-    test "Cluster is writable" {
+    test "Cluster is writable again" {
         for {set i 0} {$i < 100} {incr i} {
-            catch {$cluster set $key:$i $val:$i} err
+            catch {$cluster set $key:$i:2 $val:$i:2} err
             assert {$err eq {OK}}
         }
     }
@@ -77,7 +89,7 @@ while {[incr iterations -1]} {
     test "Restarting node #$tokill" {
         restart_instance redis $tokill
     }
-    
+
     test "Instance #$tokill is now a slave" {
         wait_for_condition 1000 50 {
             [RI $tokill role] eq {slave}
@@ -90,6 +102,8 @@ while {[incr iterations -1]} {
         for {set i 0} {$i < 100} {incr i} {
             catch {$cluster get $key:$i} err
             assert {$err eq "$val:$i"}
+            catch {$cluster get $key:$i:2} err
+            assert {$err eq "$val:$i:2"}
         }
     }
 }

tests/cluster/tests/04-resharding.tcl

@@ -0,0 +1,93 @@
+# Failover stress test.
+# In this test a different node is killed in a loop for N
+# iterations. The test checks that certain properties
+# are preseved across iterations.
+
+source "../tests/includes/init-tests.tcl"
+
+test "Create a 5 nodes cluster" {
+    create_cluster 5 5
+}
+
+test "Cluster is up" {
+    assert_cluster_state ok
+}
+
+# Return nno-zero if the specified PID is about a process still in execution,
+# otherwise 0 is returned.
+proc process_is_running {pid} {
+    # PS should return with an error if PID is non existing,
+    # and catch will return non-zero. We want to return non-zero if
+    # the PID exists, so we invert the return value with expr not operator.
+    expr {![catch {exec ps -p $pid}]}
+}
+
+# Our resharding test performs the following actions:
+#
+# - N commands are sent to the cluster in the course of the test.
+# - Every command selects a random key from key:0 to key:MAX-1.
+# - The operation RPUSH key <randomvalue> is perforemd.
+# - Tcl remembers into an array all the values pushed to each list.
+# - After N/2 commands, the resharding process is started in background.
+# - The test continues while the resharding is in progress.
+# - At the end of the test, we wait for the resharding process to stop.
+# - Finally the keys are checked to see if they contain the value they should.
+
+set numkeys 50000
+set numops 200000
+set cluster [redis_cluster 127.0.0.1:[get_instance_attrib redis 0 port]]
+catch {unset content}
+array set content {}
+set tribpid {}
+
+test "Cluster consistency during live resharding" {
+    for {set j 0} {$j < $numops} {incr j} {
+        # Trigger the resharding once we execute half the ops.
+        if {$tribpid ne {} &&
+            ($j % 10000) == 0 &&
+            ![process_is_running $tribpid]} {
+            set tribpid {}
+        }
+
+        if {$j >= $numops/2 && $tribpid eq {}} {
+            puts -nonewline "...Starting resharding..."
+            flush stdout
+            set target [dict get [get_myself [randomInt 5]] id]
+            set tribpid [lindex [exec \
+                ../../../src/redis-trib.rb reshard \
+                --from all \
+                --to $target \
+                --slots 100 \
+                --yes \
+                127.0.0.1:[get_instance_attrib redis 0 port] \
+                | [info nameofexecutable] \
+                ../tests/helpers/onlydots.tcl \
+                &] 0]
+        }
+
+        # Write random data to random list.
+        set key "key:[randomInt $numkeys]"
+        set ele [randomValue]
+        $cluster rpush $key $ele
+        lappend content($key) $ele
+
+        if {($j % 1000) == 0} {
+            puts -nonewline W; flush stdout
+        }
+    }
+
+    # Wait for the resharding process to end
+    wait_for_condition 1000 500 {
+        [process_is_running $tribpid] == 0
+    } else {
+        fail "Resharding is not terminating after some time."
+    }
+
+}
+
+test "Verify $numkeys keys for consistency with logical content" {
+    # Check that the Redis Cluster content matches our logical content.
+    foreach {key value} [array get content] {
+        assert {[$cluster lrange $key 0 -1] eq $value}
+    }
+}

tests/cluster/tests/05-slave-selection.tcl

@@ -0,0 +1,94 @@
+# Slave selection test
+# Check the algorithm trying to pick the slave with the most complete history.
+
+source "../tests/includes/init-tests.tcl"
+
+# Create a cluster with 5 master and 10 slaves, so that we have 2
+# slaves for each master.
+test "Create a 5 nodes cluster" {
+    create_cluster 5 10
+}
+
+test "Cluster is up" {
+    assert_cluster_state ok
+}
+
+test "The first master has actually two slaves" {
+    assert {[llength [lindex [R 0 role] 2]] == 2}
+}
+
+test {Slaves of #0 are instance #5 and #10 as expected} {
+    set port0 [get_instance_attrib redis 0 port]
+    assert {[lindex [R 5 role] 2] == $port0}
+    assert {[lindex [R 10 role] 2] == $port0}
+}
+
+test "Instance #5 and #10 synced with the master" {
+    wait_for_condition 1000 50 {
+        [RI 5 master_link_status] eq {up} &&
+        [RI 10 master_link_status] eq {up}
+    } else {
+        fail "Instance #5 or #10 master link status is not up"
+    }
+}
+
+set cluster [redis_cluster 127.0.0.1:[get_instance_attrib redis 0 port]]
+
+test "Slaves are both able to receive and acknowledge writes" {
+    for {set j 0} {$j < 100} {incr j} {
+        $cluster set $j $j
+    }
+    assert {[R 0 wait 2 60000] == 2}
+}
+
+test "Write data while slave #10 is paused and can't receive it" {
+    # Stop the slave with a multi/exec transaction so that the master will
+    # be killed as soon as it can accept writes again.
+    R 10 multi
+    R 10 debug sleep 10
+    R 10 client kill 127.0.0.1:$port0
+    R 10 deferred 1
+    R 10 exec
+
+    # Write some data the slave can't receive.
+    for {set j 0} {$j < 100} {incr j} {
+        $cluster set $j $j
+    }
+
+    # Prevent the master from accepting new slaves.
+    # Use a large pause value since we'll kill it anyway.
+    R 0 CLIENT PAUSE 60000
+
+    # Wait for the slave to return available again
+    R 10 deferred 0
+    assert {[R 10 read] eq {OK OK}}
+
+    # Kill the master so that a reconnection will not be possible.
+    kill_instance redis 0
+}
+
+test "Wait for instance #5 (and not #10) to turn into a master" {
+    wait_for_condition 1000 50 {
+        [RI 5 role] eq {master}
+    } else {
+        fail "No failover detected"
+    }
+}
+
+test "Wait for the node #10 to return alive before ending the test" {
+    R 10 ping
+}
+
+test "Cluster should eventually be up again" {
+    assert_cluster_state ok
+}
+
+test "Node #10 should eventually replicate node #5" {
+    set port5 [get_instance_attrib redis 5 port]
+    wait_for_condition 1000 50 {
+        ([lindex [R 10 role] 2] == $port5) &&
+        ([lindex [R 10 role] 3] eq {connected})
+    } else {
+        fail "#10 didn't became slave of #5"
+    }
+}

tests/cluster/tests/06-slave-stop-cond.tcl

@@ -0,0 +1,73 @@
+# Slave stop condition test
+# Check that if there is a disconnection time limit, the slave will not try
+# to failover its master.
+
+source "../tests/includes/init-tests.tcl"
+
+# Create a cluster with 5 master and 5 slaves.
+test "Create a 5 nodes cluster" {
+    create_cluster 5 5
+}
+
+test "Cluster is up" {
+    assert_cluster_state ok
+}
+
+test "The first master has actually one slave" {
+    assert {[llength [lindex [R 0 role] 2]] == 1}
+}
+
+test {Slaves of #0 is instance #5 as expected} {
+    set port0 [get_instance_attrib redis 0 port]
+    assert {[lindex [R 5 role] 2] == $port0}
+}
+
+test "Instance #5 synced with the master" {
+    wait_for_condition 1000 50 {
+        [RI 5 master_link_status] eq {up}
+    } else {
+        fail "Instance #5 master link status is not up"
+    }
+}
+
+test "Lower the slave validity factor of #5 to the value of 2" {
+    assert {[R 5 config set cluster-slave-validity-factor 2] eq {OK}}
+}
+
+test "Break master-slave link and prevent further reconnections" {
+    # Stop the slave with a multi/exec transaction so that the master will
+    # be killed as soon as it can accept writes again.
+    R 5 multi
+    R 5 client kill 127.0.0.1:$port0
+    # here we should sleep 6 or more seconds (node_timeout * slave_validity)
+    # but the actual validity time is actually incremented by the
+    # repl-ping-slave-period value which is 10 seconds by default. So we
+    # need to wait more than 16 seconds.
+    R 5 debug sleep 20
+    R 5 deferred 1
+    R 5 exec
+
+    # Prevent the master from accepting new slaves.
+    # Use a large pause value since we'll kill it anyway.
+    R 0 CLIENT PAUSE 60000
+
+    # Wait for the slave to return available again
+    R 5 deferred 0
+    assert {[R 5 read] eq {OK OK}}
+
+    # Kill the master so that a reconnection will not be possible.
+    kill_instance redis 0
+}
+
+test "Slave #5 is reachable and alive" {
+    assert {[R 5 ping] eq {PONG}}
+}
+
+test "Slave #5 should not be able to failover" {
+    after 10000
+    assert {[RI 5 role] eq {slave}}
+}
+
+test "Cluster should be down" {
+    assert_cluster_state fail
+}

tests/cluster/tests/07-replica-migration.tcl

@@ -0,0 +1,47 @@
+# Replica migration test.
+# Check that orphaned masters are joined by replicas of masters having
+# multiple replicas attached, according to the migration barrier settings.
+
+source "../tests/includes/init-tests.tcl"
+
+# Create a cluster with 5 master and 10 slaves, so that we have 2
+# slaves for each master.
+test "Create a 5 nodes cluster" {
+    create_cluster 5 10
+}
+
+test "Cluster is up" {
+    assert_cluster_state ok
+}
+
+test "Each master should have two replicas attached" {
+    foreach_redis_id id {
+        if {$id < 5} {
+            wait_for_condition 1000 50 {
+                [llength [lindex [R 0 role] 2]] == 2
+            } else {
+                fail "Master #$id does not have 2 slaves as expected"
+            }
+        }
+    }
+}
+
+test "Killing all the slaves of master #0 and #1" {
+    kill_instance redis 5
+    kill_instance redis 10
+    kill_instance redis 6
+    kill_instance redis 11
+    after 4000
+}
+
+foreach_redis_id id {
+    if {$id < 5} {
+        test "Master #$id should have at least one replica" {
+            wait_for_condition 1000 50 {
+                [llength [lindex [R $id role] 2]] >= 1
+            } else {
+                fail "Master #$id has no replicas"
+            }
+        }
+    }
+}

tests/cluster/tests/08-update-msg.tcl

@@ -0,0 +1,90 @@
+# Test UPDATE messages sent by other nodes when the currently authorirative
+# master is unavaialble. The test is performed in the following steps:
+#
+# 1) Master goes down.
+# 2) Slave failover and becomes new master.
+# 3) New master is partitoned away.
+# 4) Old master returns.
+# 5) At this point we expect the old master to turn into a slave ASAP because
+#    of the UPDATE messages it will receive from the other nodes when its
+#    configuration will be found to be outdated.
+
+source "../tests/includes/init-tests.tcl"
+
+test "Create a 5 nodes cluster" {
+    create_cluster 5 5
+}
+
+test "Cluster is up" {
+    assert_cluster_state ok
+}
+
+test "Cluster is writable" {
+    cluster_write_test 0
+}
+
+test "Instance #5 is a slave" {
+    assert {[RI 5 role] eq {slave}}
+}
+
+test "Instance #5 synced with the master" {
+    wait_for_condition 1000 50 {
+        [RI 5 master_link_status] eq {up}
+    } else {
+        fail "Instance #5 master link status is not up"
+    }
+}
+
+set current_epoch [CI 1 cluster_current_epoch]
+
+test "Killing one master node" {
+    kill_instance redis 0
+}
+
+test "Wait for failover" {
+    wait_for_condition 1000 50 {
+        [CI 1 cluster_current_epoch] > $current_epoch
+    } else {
+        fail "No failover detected"
+    }
+}
+
+test "Cluster should eventually be up again" {
+    assert_cluster_state ok
+}
+
+test "Cluster is writable" {
+    cluster_write_test 1
+}
+
+test "Instance #5 is now a master" {
+    assert {[RI 5 role] eq {master}}
+}
+
+test "Killing the new master #5" {
+    kill_instance redis 5
+}
+
+test "Cluster should be down now" {
+    assert_cluster_state fail
+}
+
+test "Restarting the old master node" {
+    restart_instance redis 0
+}
+
+test "Instance #0 gets converted into a slave" {
+    wait_for_condition 1000 50 {
+        [RI 0 role] eq {slave}
+    } else {
+        fail "Old master was not converted into slave"
+    }
+}
+
+test "Restarting the new master node" {
+    restart_instance redis 5
+}
+
+test "Cluster is up again" {
+    assert_cluster_state ok
+}

tests/cluster/tests/09-pubsub.tcl

@@ -0,0 +1,40 @@
+# Test PUBLISH propagation across the cluster.
+
+source "../tests/includes/init-tests.tcl"
+
+test "Create a 5 nodes cluster" {
+    create_cluster 5 5
+}
+
+proc test_cluster_publish {instance instances} {
+    # Subscribe all the instances but the one we use to send.
+    for {set j 0} {$j < $instances} {incr j} {
+        if {$j != $instance} {
+            R $j deferred 1
+            R $j subscribe testchannel
+            R $j read; # Read the subscribe reply
+        }
+    }
+
+    set data [randomValue]
+    R $instance PUBLISH testchannel $data
+
+    # Read the message back from all the nodes.
+    for {set j 0} {$j < $instances} {incr j} {
+        if {$j != $instance} {
+            set msg [R $j read]
+            assert {$data eq [lindex $msg 2]}
+            R $j unsubscribe testchannel
+            R $j read; # Read the unsubscribe reply
+            R $j deferred 0
+        }
+    }
+}
+
+test "Test publishing to master" {
+    test_cluster_publish 0 10
+}
+
+test "Test publishing to slave" {
+    test_cluster_publish 5 10
+}

tests/cluster/tests/helpers/onlydots.tcl

@@ -0,0 +1,16 @@
+# Read the standard input and only shows dots in the output, filtering out
+# all the other characters. Designed to avoid bufferization so that when
+# we get the output of redis-trib and want to show just the dots, we'll see
+# the dots as soon as redis-trib will output them.
+
+fconfigure stdin -buffering none
+
+while 1 {
+    set c [read stdin 1]
+    if {$c eq {}} {
+        exit 0; # EOF
+    } elseif {$c eq {.}} {
+        puts -nonewline .
+        flush stdout
+    }
+}

tests/cluster/tests/includes/init-tests.tcl

@@ -14,8 +14,14 @@ test "(init) Restart killed instances" {
 
 test "Cluster nodes are reachable" {
     foreach_redis_id id {
-        # Every node should just know itself.
-        assert {[R $id ping] eq {PONG}}
+        # Every node should be reachable.
+        wait_for_condition 1000 50 {
+            ([catch {R $id ping} ping_reply] == 0) &&
+            ($ping_reply eq {PONG})
+        } else {
+            catch {R $id ping} err
+            fail "Node #$id keeps replying '$err' to PING."
+        }
     }
 }
 
@@ -25,6 +31,7 @@ test "Cluster nodes hard reset" {
         R $id cluster reset hard
         R $id cluster set-config-epoch [expr {$id+1}]
         R $id config set cluster-node-timeout 3000
+        R $id config set cluster-slave-validity-factor 10
         R $id config rewrite
     }
 }

tests/instances.tcl

@@ -4,7 +4,7 @@
 # instances.
 #
 # Copyright (C) 2014 Salvatore Sanfilippo antirez@gmail.com
-# This softare is released under the BSD License. See the COPYING file for
+# This software is released under the BSD License. See the COPYING file for
 # more information.
 
 package require Tcl 8.5

tests/integration/aof.tcl

@@ -23,6 +23,84 @@ proc start_server_aof {overrides code} {
 }
 
 tags {"aof"} {
+    ## Server can start when aof-load-truncated is set to yes and AOF
+    ## is truncated, with an incomplete MULTI block.
+    create_aof {
+        append_to_aof [formatCommand set foo hello]
+        append_to_aof [formatCommand multi]
+        append_to_aof [formatCommand set bar world]
+    }
+
+    start_server_aof [list dir $server_path aof-load-truncated yes] {
+        test "Unfinished MULTI: Server should start if load-truncated is yes" {
+            assert_equal 1 [is_alive $srv]
+        }
+    }
+
+    ## Should also start with truncated AOF without incomplete MULTI block.
+    create_aof {
+        append_to_aof [formatCommand incr foo]
+        append_to_aof [formatCommand incr foo]
+        append_to_aof [formatCommand incr foo]
+        append_to_aof [formatCommand incr foo]
+        append_to_aof [formatCommand incr foo]
+        append_to_aof [string range [formatCommand incr foo] 0 end-1]
+    }
+
+    start_server_aof [list dir $server_path aof-load-truncated yes] {
+        test "Short read: Server should start if load-truncated is yes" {
+            assert_equal 1 [is_alive $srv]
+        }
+
+        set client [redis [dict get $srv host] [dict get $srv port]]
+
+        test "Truncated AOF loaded: we expect foo to be equal to 5" {
+            assert {[$client get foo] eq "5"}
+        }
+
+        test "Append a new command after loading an incomplete AOF" {
+            $client incr foo
+        }
+    }
+
+    # Now the AOF file is expected to be correct
+    start_server_aof [list dir $server_path aof-load-truncated yes] {
+        test "Short read + command: Server should start" {
+            assert_equal 1 [is_alive $srv]
+        }
+
+        set client [redis [dict get $srv host] [dict get $srv port]]
+
+        test "Truncated AOF loaded: we expect foo to be equal to 6 now" {
+            assert {[$client get foo] eq "6"}
+        }
+    }
+
+    ## Test that the server exits when the AOF contains a format error
+    create_aof {
+        append_to_aof [formatCommand set foo hello]
+        append_to_aof "!!!"
+        append_to_aof [formatCommand set foo hello]
+    }
+
+    start_server_aof [list dir $server_path aof-load-truncated yes] {
+        test "Bad format: Server should have logged an error" {
+            set pattern "*Bad file format reading the append only file*"
+            set retry 10
+            while {$retry} {
+                set result [exec tail -n1 < [dict get $srv stdout]]
+                if {[string match $pattern $result]} {
+                    break
+                }
+                incr retry -1
+                after 1000
+            }
+            if {$retry == 0} {
+                error "assertion:expected error not found on config file"
+            }
+        }
+    }
+
     ## Test the server doesn't start when the AOF contains an unfinished MULTI
     create_aof {
         append_to_aof [formatCommand set foo hello]
@@ -30,7 +108,7 @@ tags {"aof"} {
         append_to_aof [formatCommand set bar world]
     }
 
-    start_server_aof [list dir $server_path] {
+    start_server_aof [list dir $server_path aof-load-truncated no] {
         test "Unfinished MULTI: Server should have logged an error" {
             set pattern "*Unexpected end of file reading the append only file*"
             set retry 10
@@ -54,9 +132,9 @@ tags {"aof"} {
         append_to_aof [string range [formatCommand set bar world] 0 end-1]
     }
 
-    start_server_aof [list dir $server_path] {
+    start_server_aof [list dir $server_path aof-load-truncated no] {
         test "Short read: Server should have logged an error" {
-            set pattern "*Bad file format reading the append only file*"
+            set pattern "*Unexpected end of file reading the append only file*"
             set retry 10
             while {$retry} {
                 set result [exec tail -n1 < [dict get $srv stdout]]
@@ -86,12 +164,12 @@ tags {"aof"} {
     }
 
     ## Test that the server can be started using the truncated AOF
-    start_server_aof [list dir $server_path] {
+    start_server_aof [list dir $server_path aof-load-truncated no] {
         test "Fixed AOF: Server should have been started" {
             assert_equal 1 [is_alive $srv]
         }
 
-        test "Fixed AOF: Keyspace should contain values that were parsable" {   
+        test "Fixed AOF: Keyspace should contain values that were parseable" {
             set client [redis [dict get $srv host] [dict get $srv port]]
             wait_for_condition 50 100 {
                 [catch {$client ping} e] == 0
@@ -110,7 +188,7 @@ tags {"aof"} {
         append_to_aof [formatCommand spop set]
     }
 
-    start_server_aof [list dir $server_path] {
+    start_server_aof [list dir $server_path aof-load-truncated no] {
         test "AOF+SPOP: Server should have been started" {
             assert_equal 1 [is_alive $srv]
         }
@@ -133,7 +211,7 @@ tags {"aof"} {
         append_to_aof [formatCommand rpush list bar]
     }
 
-    start_server_aof [list dir $server_path] {
+    start_server_aof [list dir $server_path aof-load-truncated no] {
         test "AOF+EXPIRE: Server should have been started" {
             assert_equal 1 [is_alive $srv]
         }

tests/integration/replication.tcl

@@ -35,17 +35,17 @@ start_server {tags {"repl"}} {
 
 start_server {tags {"repl"}} {
     r set mykey foo
-    
+
     start_server {} {
         test {Second server should have role master at first} {
             s role
         } {master}
-        
+
         test {SLAVEOF should start with link status "down"} {
             r slaveof [srv -1 host] [srv -1 port]
             s master_link_status
         } {down}
-        
+
         test {The role should immediately be changed to "slave"} {
             s role
         } {slave}
@@ -54,11 +54,11 @@ start_server {tags {"repl"}} {
         test {Sync should have transferred keys from master} {
             r get mykey
         } {foo}
-        
+
         test {The link status should be up} {
             s master_link_status
         } {up}
-        
+
         test {SET on the master should immediately propagate} {
             r -1 set mykey bar
 
@@ -94,88 +94,86 @@ start_server {tags {"repl"}} {
     }
 }
 
-proc start_write_load {host port seconds} {
-    set tclsh [info nameofexecutable]
-    exec $tclsh tests/helpers/gen_write_load.tcl $host $port $seconds &
-}
-
-proc stop_write_load {handle} {
-    catch {exec /bin/kill -9 $handle}
-}
-
-start_server {tags {"repl"}} {
-    set master [srv 0 client]
-    set master_host [srv 0 host]
-    set master_port [srv 0 port]
-    set slaves {}
-    set load_handle0 [start_write_load $master_host $master_port 3]
-    set load_handle1 [start_write_load $master_host $master_port 5]
-    set load_handle2 [start_write_load $master_host $master_port 20]
-    set load_handle3 [start_write_load $master_host $master_port 8]
-    set load_handle4 [start_write_load $master_host $master_port 4]
-    start_server {} {
-        lappend slaves [srv 0 client]
+foreach dl {no yes} {
+    start_server {tags {"repl"}} {
+        set master [srv 0 client]
+        $master config set repl-diskless-sync $dl
+        set master_host [srv 0 host]
+        set master_port [srv 0 port]
+        set slaves {}
+        set load_handle0 [start_write_load $master_host $master_port 3]
+        set load_handle1 [start_write_load $master_host $master_port 5]
+        set load_handle2 [start_write_load $master_host $master_port 20]
+        set load_handle3 [start_write_load $master_host $master_port 8]
+        set load_handle4 [start_write_load $master_host $master_port 4]
         start_server {} {
             lappend slaves [srv 0 client]
             start_server {} {
                 lappend slaves [srv 0 client]
-                test "Connect multiple slaves at the same time (issue #141)" {
-                    # Send SALVEOF commands to slaves
-                    [lindex $slaves 0] slaveof $master_host $master_port
-                    [lindex $slaves 1] slaveof $master_host $master_port
-                    [lindex $slaves 2] slaveof $master_host $master_port
+                start_server {} {
+                    lappend slaves [srv 0 client]
+                    test "Connect multiple slaves at the same time (issue #141), diskless=$dl" {
+                        # Send SALVEOF commands to slaves
+                        [lindex $slaves 0] slaveof $master_host $master_port
+                        [lindex $slaves 1] slaveof $master_host $master_port
+                        [lindex $slaves 2] slaveof $master_host $master_port
 
-                    # Wait for all the three slaves to reach the "online" state
-                    set retry 500
-                    while {$retry} {
-                        set info [r -3 info]
-                        if {[string match {*slave0:*state=online*slave1:*state=online*slave2:*state=online*} $info]} {
-                            break
-                        } else {
-                            incr retry -1
-                            after 100
+                        # Wait for all the three slaves to reach the "online"
+                        # state from the POV of the master.
+                        set retry 500
+                        while {$retry} {
+                            set info [r -3 info]
+                            if {[string match {*slave0:*state=online*slave1:*state=online*slave2:*state=online*} $info]} {
+                                break
+                            } else {
+                                incr retry -1
+                                after 100
+                            }
+                        }
+                        if {$retry == 0} {
+                            error "assertion:Slaves not correctly synchronized"
                         }
-                    }
-                    if {$retry == 0} {
-                        error "assertion:Slaves not correctly synchronized"
-                    }
 
-                    # Stop the write load
-                    stop_write_load $load_handle0
-                    stop_write_load $load_handle1
-                    stop_write_load $load_handle2
-                    stop_write_load $load_handle3
-                    stop_write_load $load_handle4
+                        # Wait that slaves acknowledge they are online so
+                        # we are sure that DBSIZE and DEBUG DIGEST will not
+                        # fail because of timing issues.
+                        wait_for_condition 500 100 {
+                            [lindex [[lindex $slaves 0] role] 3] eq {connected} &&
+                            [lindex [[lindex $slaves 1] role] 3] eq {connected} &&
+                            [lindex [[lindex $slaves 2] role] 3] eq {connected}
+                        } else {
+                            fail "Slaves still not connected after some time"
+                        }
 
-                    # Wait that slaves exit the "loading" state
-                    wait_for_condition 500 100 {
-                        ![string match {*loading:1*} [[lindex $slaves 0] info]] &&
-                        ![string match {*loading:1*} [[lindex $slaves 1] info]] &&
-                        ![string match {*loading:1*} [[lindex $slaves 2] info]]
-                    } else {
-                        fail "Slaves still loading data after too much time"
+                        # Stop the write load
+                        stop_write_load $load_handle0
+                        stop_write_load $load_handle1
+                        stop_write_load $load_handle2
+                        stop_write_load $load_handle3
+                        stop_write_load $load_handle4
+
+                        # Make sure that slaves and master have same
+                        # number of keys
+                        wait_for_condition 500 100 {
+                            [$master dbsize] == [[lindex $slaves 0] dbsize] &&
+                            [$master dbsize] == [[lindex $slaves 1] dbsize] &&
+                            [$master dbsize] == [[lindex $slaves 2] dbsize]
+                        } else {
+                            fail "Different number of keys between masted and slave after too long time."
+                        }
+
+                        # Check digests
+                        set digest [$master debug digest]
+                        set digest0 [[lindex $slaves 0] debug digest]
+                        set digest1 [[lindex $slaves 1] debug digest]
+                        set digest2 [[lindex $slaves 2] debug digest]
+                        assert {$digest ne 0000000000000000000000000000000000000000}
+                        assert {$digest eq $digest0}
+                        assert {$digest eq $digest1}
+                        assert {$digest eq $digest2}
                     }
-
-                    # Make sure that slaves and master have same number of keys
-                    wait_for_condition 500 100 {
-                        [$master dbsize] == [[lindex $slaves 0] dbsize] &&
-                        [$master dbsize] == [[lindex $slaves 1] dbsize] &&
-                        [$master dbsize] == [[lindex $slaves 2] dbsize]
-                    } else {
-                        fail "Different number of keys between masted and slave after too long time."
-                    }
-
-                    # Check digests
-                    set digest [$master debug digest]
-                    set digest0 [[lindex $slaves 0] debug digest]
-                    set digest1 [[lindex $slaves 1] debug digest]
-                    set digest2 [[lindex $slaves 2] debug digest]
-                    assert {$digest ne 0000000000000000000000000000000000000000}
-                    assert {$digest eq $digest0}
-                    assert {$digest eq $digest1}
-                    assert {$digest eq $digest2}
-                }
-           }
+               }
+            }
         }
     }
 }

tests/sentinel/run.tcl

@@ -1,5 +1,5 @@
 # Sentinel test suite. Copyright (C) 2014 Salvatore Sanfilippo antirez@gmail.com
-# This softare is released under the BSD License. See the COPYING file for
+# This software is released under the BSD License. See the COPYING file for
 # more information.
 
 cd tests/sentinel

tests/support/redis.tcl

@@ -18,7 +18,7 @@
 #         $r ping [list handlePong]
 #     }
 # }
-# 
+#
 # set r [redis]
 # $r blocking 0
 # $r get fo [list handlePong]

tests/support/server.tcl

@@ -70,6 +70,9 @@ proc kill_server config {
     if {$::valgrind} {
         check_valgrind_errors [dict get $config stderr]
     }
+
+    # Remove this pid from the set of active pids in the test server.
+    send_data_packet $::test_server_fd server-killed $pid
 }
 
 proc is_alive config {
@@ -178,10 +181,10 @@ proc start_server {options {code undefined}} {
             dict set config $directive $arguments
         }
     }
-    
+
     # use a different directory every time a server is started
     dict set config dir [tmpdir server]
-    
+
     # start every server on a different port
     set ::port [find_available_port [expr {$::port+1}]]
     dict set config port $::port
@@ -190,7 +193,7 @@ proc start_server {options {code undefined}} {
     foreach {directive arguments} [concat $::global_overrides $overrides] {
         dict set config $directive $arguments
     }
-    
+
     # write new configuration to temporary file
     set config_file [tmpfile redis.conf]
     set fp [open $config_file w+]
@@ -204,11 +207,14 @@ proc start_server {options {code undefined}} {
     set stderr [format "%s/%s" [dict get $config "dir"] "stderr"]
 
     if {$::valgrind} {
-        exec valgrind --suppressions=src/valgrind.sup --show-reachable=no --show-possibly-lost=no --leak-check=full src/redis-server $config_file > $stdout 2> $stderr &
+        set pid [exec valgrind --suppressions=src/valgrind.sup --show-reachable=no --show-possibly-lost=no --leak-check=full src/redis-server $config_file > $stdout 2> $stderr &]
     } else {
-        exec src/redis-server $config_file > $stdout 2> $stderr &
+        set pid [exec src/redis-server $config_file > $stdout 2> $stderr &]
     }
-    
+
+    # Tell the test server about this new instance.
+    send_data_packet $::test_server_fd server-spawned $pid
+
     # check that the server actually started
     # ugly but tries to be as fast as possible...
     if {$::valgrind} {set retrynum 1000} else {set retrynum 100}
@@ -233,10 +239,10 @@ proc start_server {options {code undefined}} {
         start_server_error $config_file $err
         return
     }
-    
-    # find out the pid
-    while {![info exists pid]} {
-        regexp {PID:\s(\d+)} [exec cat $stdout] _ pid
+
+    # Wait for actual startup
+    while {![info exists _pid]} {
+        regexp {PID:\s(\d+)} [exec cat $stdout] _ _pid
         after 100
     }
 

tests/support/test.tcl

@@ -29,7 +29,7 @@ proc assert_error {pattern code} {
     if {[catch {uplevel 1 $code} error]} {
         assert_match $pattern $error
     } else {
-        error "assertion:Expected an error but nothing was catched"
+        error "assertion:Expected an error but nothing was caught"
     }
 }
 

tests/support/util.tcl

@@ -353,9 +353,21 @@ proc colorstr {color str} {
             default {set colorcode {37}}
         }
         if {$colorcode ne {}} {
-            return "\033\[$b;${colorcode};40m$str\033\[0m"
+            return "\033\[$b;${colorcode};49m$str\033\[0m"
         }
     } else {
         return $str
     }
 }
+
+# Execute a background process writing random data for the specified number
+# of seconds to the specified Redis instance.
+proc start_write_load {host port seconds} {
+    set tclsh [info nameofexecutable]
+    exec $tclsh tests/helpers/gen_write_load.tcl $host $port $seconds &
+}
+
+# Stop a process generating write load executed with start_write_load.
+proc stop_write_load {handle} {
+    catch {exec /bin/kill -9 $handle}
+}

tests/test_helper.tcl

@@ -1,5 +1,5 @@
 # Redis test suite. Copyright (C) 2009 Salvatore Sanfilippo antirez@gmail.com
-# This softare is released under the BSD License. See the COPYING file for
+# This software is released under the BSD License. See the COPYING file for
 # more information.
 
 package require Tcl 8.5
@@ -65,6 +65,9 @@ set ::file ""; # If set, runs only the tests in this comma separated list
 set ::curfile ""; # Hold the filename of the current suite
 set ::accurate 0; # If true runs fuzz tests with more iterations
 set ::force_failure 0
+set ::timeout 600; # 10 minutes without progresses will quit the test.
+set ::last_progress [clock seconds]
+set ::active_servers {} ; # Pids of active Redis instances.
 
 # Set to 1 when we are running in client mode. The Redis test uses a
 # server-client model to run tests simultaneously. The server instance
@@ -200,11 +203,19 @@ proc test_server_main {} {
     vwait forever
 }
 
-# This function gets called 10 times per second, for now does nothing but
-# may be used in the future in order to detect test clients taking too much
-# time to execute the task.
+# This function gets called 10 times per second.
 proc test_server_cron {} {
-    # Do some work here.
+    set elapsed [expr {[clock seconds]-$::last_progress}]
+
+    if {$elapsed > $::timeout} {
+        set err "\[[colorstr red TIMEOUT]\]: clients state report follows."
+        puts $err
+        show_clients_state
+        kill_clients
+        force_kill_all_servers
+        the_end
+    }
+
     after 100 test_server_cron
 }
 
@@ -230,6 +241,8 @@ proc read_from_test_client fd {
     set bytes [gets $fd]
     set payload [read $fd $bytes]
     foreach {status data} $payload break
+    set ::last_progress [clock seconds]
+
     if {$status eq {ready}} {
         if {!$::quiet} {
             puts "\[$status\]: $data"
@@ -256,12 +269,15 @@ proc read_from_test_client fd {
         set ::active_clients_task($fd) "(ERR) $data"
     } elseif {$status eq {exception}} {
         puts "\[[colorstr red $status]\]: $data"
-        foreach p $::clients_pids {
-            catch {exec kill -9 $p}
-        }
+        kill_clients
+        force_kill_all_servers
         exit 1
     } elseif {$status eq {testing}} {
         set ::active_clients_task($fd) "(IN PROGRESS) $data"
+    } elseif {$status eq {server-spawned}} {
+        lappend ::active_servers $data
+    } elseif {$status eq {server-killed}} {
+        set ::active_servers [lsearch -all -inline -not -exact $::active_servers $data]
     } else {
         if {!$::quiet} {
             puts "\[$status\]: $data"
@@ -269,6 +285,31 @@ proc read_from_test_client fd {
     }
 }
 
+proc show_clients_state {} {
+    # The following loop is only useful for debugging tests that may
+    # enter an infinite loop. Commented out normally.
+    foreach x $::active_clients {
+        if {[info exist ::active_clients_task($x)]} {
+            puts "$x => $::active_clients_task($x)"
+        } else {
+            puts "$x => ???"
+        }
+    }
+}
+
+proc kill_clients {} {
+    foreach p $::clients_pids {
+        catch {exec kill $p}
+    }
+}
+
+proc force_kill_all_servers {} {
+    foreach p $::active_servers {
+        puts "Killing still running Redis server $p"
+        catch {exec kill -9 $p}
+    }
+}
+
 # A new client is idle. Remove it from the list of active clients and
 # if there are still test units to run, launch them.
 proc signal_idle_client fd {
@@ -276,17 +317,7 @@ proc signal_idle_client fd {
     set ::active_clients \
         [lsearch -all -inline -not -exact $::active_clients $fd]
 
-    if 0 {
-        # The following loop is only useful for debugging tests that may
-        # enter an infinite loop. Commented out normally.
-        foreach x $::active_clients {
-            if {[info exist ::active_clients_task($x)]} {
-                puts "$x => $::active_clients_task($x)"
-            } else {
-                puts "$x => ???"
-            }
-        }
-    }
+    if 0 {show_clients_state}
 
     # New unit to process?
     if {$::next_test != [llength $::all_tests]} {
@@ -306,7 +337,7 @@ proc signal_idle_client fd {
     }
 }
 
-# The the_end funciton gets called when all the test units were already
+# The the_end function gets called when all the test units were already
 # executed, so the test finished.
 proc the_end {} {
     # TODO: print the status, exit with the rigth exit code.
@@ -361,7 +392,8 @@ proc print_help_screen {} {
         "--quiet            Don't show individual tests."
         "--single <unit>    Just execute the specified unit (see next option)."
         "--list-tests       List all the available test units."
-        "--clients <num>    Number of test clients (16)."
+        "--clients <num>    Number of test clients (default 16)."
+        "--timeout <sec>    Test timeout in seconds (default 10 min)."
         "--force-failure    Force the execution of a test that always fails."
         "--help             Print this help screen."
     } "\n"]
@@ -410,6 +442,9 @@ for {set j 0} {$j < [llength $argv]} {incr j} {
     } elseif {$opt eq {--clients}} {
         set ::numclients $arg
         incr j
+    } elseif {$opt eq {--timeout}} {
+        set ::timeout $arg
+        incr j
     } elseif {$opt eq {--help}} {
         print_help_screen
         exit 0

tests/unit/aofrw.tcl

@@ -1,5 +1,67 @@
 start_server {tags {"aofrw"}} {
+    # Enable the AOF
+    r config set appendonly yes
+    r config set auto-aof-rewrite-percentage 0 ; # Disable auto-rewrite.
+    waitForBgrewriteaof r
 
+    test {AOF rewrite during write load} {
+        # Start a write load for 10 seconds
+        set master [srv 0 client]
+        set master_host [srv 0 host]
+        set master_port [srv 0 port]
+        set load_handle0 [start_write_load $master_host $master_port 10]
+        set load_handle1 [start_write_load $master_host $master_port 10]
+        set load_handle2 [start_write_load $master_host $master_port 10]
+        set load_handle3 [start_write_load $master_host $master_port 10]
+        set load_handle4 [start_write_load $master_host $master_port 10]
+
+        # Make sure the instance is really receiving data
+        wait_for_condition 50 100 {
+            [r dbsize] > 0
+        } else {
+            fail "No write load detected."
+        }
+
+        # After 3 seconds, start a rewrite, while the write load is still
+        # active.
+        after 3000
+        r bgrewriteaof
+        waitForBgrewriteaof r
+
+        # Let it run a bit more so that we'll append some data to the new
+        # AOF.
+        after 1000
+
+        # Stop the processes generating the load if they are still active
+        stop_write_load $load_handle0
+        stop_write_load $load_handle1
+        stop_write_load $load_handle2
+        stop_write_load $load_handle3
+        stop_write_load $load_handle4
+
+        # Make sure that we remain the only connected client.
+        # This step is needed to make sure there are no pending writes
+        # that will be processed between the two "debug digest" calls.
+        wait_for_condition 50 100 {
+            [llength [split [string trim [r client list]] "\n"]] == 1
+        } else {
+            puts [r client list]
+            fail "Clients generating loads are not disconnecting"
+        }
+
+        # Get the data set digest
+        set d1 [r debug digest]
+
+        # Load the AOF
+        r debug loadaof
+        set d2 [r debug digest]
+
+        # Make sure they are the same
+        assert {$d1 eq $d2}
+    }
+}
+
+start_server {tags {"aofrw"}} {
     test {Turning off AOF kills the background writing child if any} {
         r config set appendonly yes
         waitForBgrewriteaof r

tests/unit/auth.tcl

@@ -10,7 +10,7 @@ start_server {tags {"auth"} overrides {requirepass foobar}} {
         catch {r auth wrong!} err
         set _ $err
     } {ERR*invalid password}
-    
+
     test {Arbitrary command gives an error when AUTH is required} {
         catch {r set foo bar} err
         set _ $err

tests/unit/basic.tcl

@@ -83,7 +83,7 @@ start_server {tags {"basic"}} {
             for {set x 9999} {$x >= 0} {incr x -1} {
                 set val [r get $x]
                 if {$val ne $x} {
-                    set err "Eleemnt at position $x is $val instead of $x"
+                    set err "Element at position $x is $val instead of $x"
                     break
                 }
             }
@@ -150,6 +150,29 @@ start_server {tags {"basic"}} {
         r decrby novar 17179869185
     } {-1}
 
+    test {INCR uses shared objects in the 0-9999 range} {
+        r set foo -1
+        r incr foo
+        assert {[r object refcount foo] > 1}
+        r set foo 9998
+        r incr foo
+        assert {[r object refcount foo] > 1}
+        r incr foo
+        assert {[r object refcount foo] == 1}
+    }
+
+    test {INCR can modify objects in-place} {
+        r set foo 20000
+        r incr foo
+        assert {[r object refcount foo] == 1}
+        set old [lindex [split [r debug object foo]] 1]
+        r incr foo
+        set new [lindex [split [r debug object foo]] 1]
+        assert {[string range $old 0 2] eq "at:"}
+        assert {[string range $new 0 2] eq "at:"}
+        assert {$old eq $new}
+    }
+
     test {INCRBYFLOAT against non existing key} {
         r del novar
         list    [roundFloat [r incrbyfloat novar 1]] \
@@ -300,7 +323,7 @@ start_server {tags {"basic"}} {
         catch {r foobaredcommand} err
         string match ERR* $err
     } {1}
-    
+
     test {RENAME basic usage} {
         r set mykey hello
         r rename mykey mykey1
@@ -404,6 +427,12 @@ start_server {tags {"basic"}} {
         r move mykey 10
     } {0}
 
+    test {MOVE against non-integer DB (#1428)} {
+        r set mykey hello
+        catch {r move mykey notanumber} e
+        set e
+    } {*ERR*index out of range}
+
     test {SET/GET keys in different DBs} {
         r set a hello
         r set b world
@@ -420,7 +449,7 @@ start_server {tags {"basic"}} {
         r select 9
         format $res
     } {hello world foo bared}
-    
+
     test {MGET} {
         r flushdb
         r set foo BAR
@@ -476,7 +505,7 @@ start_server {tags {"basic"}} {
         r set foo bar
         list [r getset foo xyz] [r get foo]
     } {bar xyz}
-    
+
     test {MSET base case} {
         r mset x 10 y "foo bar" z "x x x x x x x\n\n\r\n"
         r mget x y z
@@ -769,4 +798,9 @@ start_server {tags {"basic"}} {
         r keys *
         r keys *
     } {dlskeriewrioeuwqoirueioqwrueoqwrueqw}
+
+    test {GETRANGE with huge ranges, Github issue #1844} {
+        r set foo bar
+        r getrange foo 0 4294967297
+    } {bar}
 }

tests/unit/bitops.tcl

@@ -125,7 +125,7 @@ start_server {tags {"bitops"}} {
     test {BITOP where dest and target are the same key} {
         r set s "\xaa\x00\xff\x55"
         r bitop not s s
-        r get s 
+        r get s
     } "\x55\xff\x00\xaa"
 
     test {BITOP AND|OR|XOR don't change the string with single input key} {

tests/unit/introspection.tcl

@@ -27,7 +27,7 @@ start_server {tags {"introspection"}} {
     test {CLIENT LIST shows empty fields for unassigned names} {
         r client list
     } {*name= *}
-    
+
     test {CLIENT SETNAME does not accept spaces} {
         catch {r client setname "foo bar"} e
         set e