Solo tests should work with loopn

* added replping command for initiating replication

* backwards compatibility for replping (retry if not recognized)

* don't allow ping during loading (load balancer fix)

* changed replping warning to notice

src/Makefile

@@ -127,7 +127,7 @@ ifneq ($(uname_S),Darwin)
 endif
 # Linux ARM32 needs -latomic at linking time
 ifneq (,$(findstring armv,$(uname_M)))
-        FINAL_LIBS+=-latomic
+	FINAL_LIBS+=-latomic
 endif
 
 
@@ -177,7 +177,7 @@ ifeq ($(uname_S),OpenBSD)
 	    FINAL_CXXFLAGS+= -DUSE_BACKTRACE -I/usr/local/include
 	    FINAL_LDFLAGS+= -L/usr/local/lib
 	    FINAL_LIBS+= -lexecinfo
-    	endif
+	endif
 
 else
 ifeq ($(uname_S),NetBSD)
@@ -187,7 +187,7 @@ ifeq ($(uname_S),NetBSD)
 	    FINAL_CFLAGS+= -DUSE_BACKTRACE -I/usr/pkg/include
 	    FINAL_LDFLAGS+= -L/usr/pkg/lib
 	    FINAL_LIBS+= -lexecinfo
-    	endif
+	endif
 else
 ifeq ($(uname_S),FreeBSD)
 	# FreeBSD
@@ -316,6 +316,14 @@ else
     endef
 endif
 
+# Alpine OS doesn't have support for the execinfo backtrace library we use for debug, so we provide an alternate implementation using libwunwind.
+OS := $(shell cat /etc/os-release | grep ID= | head -n 1 | cut -d'=' -f2)
+ifeq ($(OS),alpine)
+    FINAL_CXXFLAGS+=-DUNW_LOCAL_ONLY
+    FINAL_LIBS += -lunwind
+endif
+
+
 REDIS_CC=$(QUIET_CC)$(CC) $(FINAL_CFLAGS)
 REDIS_CXX=$(QUIET_CC)$(CXX) $(FINAL_CXXFLAGS)
 KEYDB_AS=$(QUIET_CC) as --64 -g

src/debug.cpp

@@ -51,6 +51,12 @@ typedef ucontext_t sigcontext_t;
 #include <cxxabi.h>
 #endif /* HAVE_BACKTRACE */
 
+//UNW_LOCAL_ONLY being set means we use libunwind for backtraces instead of execinfo
+#ifdef UNW_LOCAL_ONLY
+#include <libunwind.h>
+#include <cxxabi.h>
+#endif
+
 #ifdef __CYGWIN__
 #ifndef SA_ONSTACK
 #define SA_ONSTACK 0x08000000
@@ -947,7 +953,7 @@ void _serverAssert(const char *estr, const char *file, int line) {
     serverLog(LL_WARNING,"==> %s:%d '%s' is not true",file,line,estr);
 
     if (g_pserver->crashlog_enabled) {
-#ifdef HAVE_BACKTRACE
+#if defined HAVE_BACKTRACE || defined UNW_LOCAL_ONLY
         logStackTrace(NULL, 1);
 #endif
         printCrashReport();
@@ -1044,7 +1050,7 @@ void _serverPanic(const char *file, int line, const char *msg, ...) {
     serverLog(LL_WARNING,"Guru Meditation: %s #%s:%d",fmtmsg,file,line);
 
     if (g_pserver->crashlog_enabled) {
-#ifdef HAVE_BACKTRACE
+#if defined HAVE_BACKTRACE || defined UNW_LOCAL_ONLY
         logStackTrace(NULL, 1);
 #endif
         printCrashReport();
@@ -1599,6 +1605,65 @@ void safe_write(int fd, const void *pv, ssize_t cb)
     } while (offset < cb);
 }
 
+#ifdef UNW_LOCAL_ONLY
+
+/* Logs the stack trace using the libunwind call.
+ * The eip argument is unused as libunwind only gets local context.
+ * The uplevel argument indicates how many of the calling functions to skip.
+ */
+void logStackTrace(void * eip, int uplevel) {
+    (void)eip;//UNUSED
+    const char *msg;
+    int fd = openDirectLogFiledes();
+
+    if (fd == -1) return; /* If we can't log there is anything to do. */
+
+    msg = "\n------ STACK TRACE ------\n";
+    if (write(fd,msg,strlen(msg)) == -1) {/* Avoid warning. */};
+    unw_cursor_t cursor;
+    unw_context_t context;
+
+    unw_getcontext(&context);
+    unw_init_local(&cursor, &context);
+
+    /* Write symbols to log file */
+    msg = "\nBacktrace:\n";
+    if (write(fd,msg,strlen(msg)) == -1) {/* Avoid warning. */};
+
+    for (int i = 0; i < uplevel; i++) {
+        unw_step(&cursor);
+    }
+
+    while ( unw_step(&cursor) ) {
+    unw_word_t ip, sp, off;
+
+    unw_get_reg(&cursor, UNW_REG_IP, &ip);
+    unw_get_reg(&cursor, UNW_REG_SP, &sp);
+
+    char symbol[256] = {"<unknown>"};
+    char *name = symbol;
+
+    if ( !unw_get_proc_name(&cursor, symbol, sizeof(symbol), &off) ) {
+        int status;
+        if ( (name = abi::__cxa_demangle(symbol, NULL, NULL, &status)) == 0 )
+        name = symbol;
+    }
+
+    dprintf(fd, "%s(+0x%" PRIxPTR ") [0x%016" PRIxPTR "] sp=0x%016" PRIxPTR "\n",
+        name,
+        static_cast<uintptr_t>(off),
+        static_cast<uintptr_t>(ip),
+        static_cast<uintptr_t>(sp));
+
+    if ( name != symbol )
+        free(name);
+    }
+}
+
+#endif /* UNW_LOCAL_ONLY */
+
+#ifdef HAVE_BACKTRACE
+
 void backtrace_symbols_demangle_fd(void **trace, size_t csym, int fd)
 {
     char **syms = backtrace_symbols(trace, csym);
@@ -1642,8 +1707,6 @@ void backtrace_symbols_demangle_fd(void **trace, size_t csym, int fd)
     free(syms);
 }
 
-#ifdef HAVE_BACKTRACE
-
 /* Logs the stack trace using the backtrace() call. This function is designed
  * to be called from signal handlers safely.
  * The eip argument is optional (can take NULL).
@@ -1934,6 +1997,9 @@ void sigsegvHandler(int sig, siginfo_t *info, void *secret) {
 
     logRegisters(uc);
 #endif
+#ifdef UNW_LOCAL_ONLY
+    logStackTrace(NULL, 1);
+#endif 
 
     printCrashReport();
 
@@ -2028,6 +2094,8 @@ void watchdogSignalHandler(int sig, siginfo_t *info, void *secret) {
     serverLogFromHandler(LL_WARNING,"\n--- WATCHDOG TIMER EXPIRED ---");
 #ifdef HAVE_BACKTRACE
     logStackTrace(getMcontextEip(uc), 1);
+#elif defined UNW_LOCAL_ONLY
+    logStackTrace(NULL, 1);
 #else
     serverLogFromHandler(LL_WARNING,"Sorry: no support for backtrace().");
 #endif

src/replication.cpp

@@ -2759,18 +2759,19 @@ void syncWithMaster(connection *conn) {
         goto error;
     }
 
+retry_connect:
     /* Send a PING to check the master is able to reply without errors. */
-    if (mi->repl_state == REPL_STATE_CONNECTING) {
+    if (mi->repl_state == REPL_STATE_CONNECTING || mi->repl_state == REPL_STATE_RETRY_NOREPLPING) {
         serverLog(LL_NOTICE,"Non blocking connect for SYNC fired the event.");
         /* Delete the writable event so that the readable event remains
          * registered and we can wait for the PONG reply. */
         connSetReadHandler(conn, syncWithMaster);
         connSetWriteHandler(conn, NULL);
-        mi->repl_state = REPL_STATE_RECEIVE_PING_REPLY;
         /* Send the PING, don't check for errors at all, we have the timeout
          * that will take care about this. */
-        err = sendCommand(conn,"PING",NULL);
+        err = sendCommand(conn,mi->repl_state == REPL_STATE_RETRY_NOREPLPING ? "PING" : "REPLPING",NULL);
         if (err) goto write_error;
+        mi->repl_state = REPL_STATE_RECEIVE_PING_REPLY;
         return;
     }
 
@@ -2783,7 +2784,13 @@ void syncWithMaster(connection *conn) {
          * Note that older versions of Redis replied with "operation not
          * permitted" instead of using a proper error code, so we test
          * both. */
-        if (err[0] != '+' &&
+        if (strncmp(err,"-ERR unknown command",20) == 0) {
+            serverLog(LL_NOTICE,"Master does not support REPLPING, sending PING instead...");
+            mi->repl_state = REPL_STATE_RETRY_NOREPLPING;
+            sdsfree(err);
+            err = NULL;
+            goto retry_connect;
+        } else if (err[0] != '+' &&
             strncmp(err,"-NOAUTH",7) != 0 &&
             strncmp(err,"-NOPERM",7) != 0 &&
             strncmp(err,"-ERR operation not permitted",28) != 0)

src/server.cpp

@@ -753,6 +753,10 @@ struct redisCommand redisCommandTable[] = {
      "ok-stale ok-loading fast @connection @replication",
      0,NULL,0,0,0,0,0,0},
 
+     {"replping",pingCommand,-1,
+     "ok-stale fast @connection @replication",
+     0,NULL,0,0,0,0,0,0},
+
     {"echo",echoCommand,2,
      "fast @connection",
      0,NULL,0,0,0,0,0,0},
@@ -2768,6 +2772,7 @@ void createSharedObjects(void) {
     shared.lastid = makeObjectShared("LASTID",6);
     shared.default_username = makeObjectShared("default",7);
     shared.ping = makeObjectShared("ping",4);
+    shared.replping = makeObjectShared("replping", 8);
     shared.setid = makeObjectShared("SETID",5);
     shared.keepttl = makeObjectShared("KEEPTTL",7);
     shared.load = makeObjectShared("LOAD",4);

src/server.h

@@ -505,6 +505,7 @@ typedef enum {
     REPL_STATE_NONE = 0,            /* No active replication */
     REPL_STATE_CONNECT,             /* Must connect to master */
     REPL_STATE_CONNECTING,          /* Connecting to master */
+    REPL_STATE_RETRY_NOREPLPING,    /* Master does not support REPLPING, retry with PING */
     /* --- Handshake states, must be ordered --- */
     REPL_STATE_RECEIVE_PING_REPLY,  /* Wait for PING reply */
     REPL_STATE_SEND_HANDSHAKE,      /* Send handshake sequance to master */
@@ -1283,7 +1284,7 @@ struct sharedObjectsStruct {
     *emptyscan, *multi, *exec, *left, *right, *hset, *srem, *xgroup, *xclaim,  
     *script, *replconf, *eval, *persist, *set, *pexpireat, *pexpire, 
     *time, *pxat, *px, *retrycount, *force, *justid, 
-    *lastid, *ping, *setid, *keepttl, *load, *createconsumer,
+    *lastid, *ping, *replping, *setid, *keepttl, *load, *createconsumer,
     *getack, *special_asterick, *special_equals, *default_username,
     *hdel, *zrem, *mvccrestore, *pexpirememberat,
     *select[PROTO_SHARED_SELECT_CMDS],

tests/test_helper.tcl

@@ -253,7 +253,7 @@ proc s {args} {
 # test server, so that the test server will send them again to
 # clients once the clients are idle.
 proc run_solo {name code} {
-    if {$::numclients == 1 || $::loop || $::external  || $::endurance} {
+    if {$::numclients == 1 || $::loop < 0 || $::external || $::endurance} {
         # run_solo is not supported in these scenarios, just run the code.
         if {$::endurance} {
             puts "Skipping solo tests because endurance mode is enabled"