Merge branch 'egil/fix-crashdump-default/OTP-10602' into maint-r15

* egil/fix-crashdump-default/OTP-10602:
  erts: Change default of erl_crash.dump

5 files changed, 60 insertions, 23 deletions

diff --git a/erts/emulator/beam/break.c b/erts/emulator/beam/break.c
index cf66f4e6b6..f7e9f15655 100644
--- a/erts/emulator/beam/break.c
+++ b/erts/emulator/beam/break.c
@@ -657,6 +657,7 @@ erl_crash_dump_v(char *file, int line, char* fmt, va_list args)
     char dumpnamebuf[MAXPATHLEN];
     char* dumpname;
     int secs;
+    int env_erl_crash_dump_seconds_set = 1;
 
     if (ERTS_SOMEONE_IS_CRASH_DUMPING)
 	return;
@@ -681,6 +682,8 @@ erl_crash_dump_v(char *file, int line, char* fmt, va_list args)
 
     envsz = sizeof(env);
     /* ERL_CRASH_DUMP_SECONDS not set
+     * if we have a heart port, break immediately
+     * otherwise dump crash indefinitely (until crash is complete)
      * same as ERL_CRASH_DUMP_SECONDS = 0
      * - do not write dump
      * - do not set an alarm
@@ -702,8 +705,10 @@ erl_crash_dump_v(char *file, int line, char* fmt, va_list args)
      */
 	
     if (erts_sys_getenv__("ERL_CRASH_DUMP_SECONDS", env, &envsz) != 0) {
-	return; /* break immediately */
+	env_erl_crash_dump_seconds_set = 0;
+	secs = -1;
     } else {
+	env_erl_crash_dump_seconds_set = 1;
 	secs = atoi(env);
     }
 
@@ -711,7 +716,16 @@ erl_crash_dump_v(char *file, int line, char* fmt, va_list args)
 	return;
     }
 
-    erts_sys_prepare_crash_dump(secs);
+    /* erts_sys_prepare_crash_dump returns 1 if heart port is found, otherwise 0
+     * If we don't find heart (0) and we don't have ERL_CRASH_DUMP_SECONDS set
+     * we should continue writing a dump
+     *
+     * beware: secs -1 means no alarm
+     */
+
+    if (erts_sys_prepare_crash_dump(secs) && !env_erl_crash_dump_seconds_set ) {
+	return;
+    }
 
     if (erts_sys_getenv__("ERL_CRASH_DUMP",&dumpnamebuf[0],&dumpnamebufsize) != 0)
 	dumpname = "erl_crash.dump";
diff --git a/erts/emulator/beam/sys.h b/erts/emulator/beam/sys.h
index 9dd8341520..41389c8734 100644
--- a/erts/emulator/beam/sys.h
+++ b/erts/emulator/beam/sys.h
@@ -652,7 +652,7 @@ void erts_sys_schedule_interrupt_timed(int set, erts_short_time_t msec);
 void erts_sys_main_thread(void);
 #endif
 
-extern void erts_sys_prepare_crash_dump(int secs);
+extern int erts_sys_prepare_crash_dump(int secs);
 extern void erts_sys_pre_init(void);
 extern void erl_sys_init(void);
 extern void erl_sys_args(int *argc, char **argv);
diff --git a/erts/emulator/sys/unix/sys.c b/erts/emulator/sys/unix/sys.c
index c485a4eece..964751cf86 100644
--- a/erts/emulator/sys/unix/sys.c
+++ b/erts/emulator/sys/unix/sys.c
@@ -686,7 +686,7 @@ static RETSIGTYPE break_handler(int sig)
 }
 #endif /* 0 */
 
-static ERTS_INLINE void
+static ERTS_INLINE int
 prepare_crash_dump(int secs)
 {
 #define NUFBUF (3)
@@ -698,19 +698,35 @@ prepare_crash_dump(int secs)
     Eterm *hp = heap;
     Eterm list = NIL;
     int heart_fd[2] = {-1,-1};
+    int has_heart = 0;
 
     UseTmpHeapNoproc(NUFBUF);
 
     if (ERTS_PREPARED_CRASH_DUMP)
-	return; /* We have already been called */
+	return 0; /* We have already been called */
 
     heart_port = erts_get_heart_port();
+
+    /* Positive secs means an alarm must be set
+     * 0 or negative means no alarm
+     *
+     * Set alarm before we try to write to a port
+     * we don't want to hang on a port write with
+     * no alarm.
+     *
+     */
+
+    if (secs >= 0) {
+	alarm((unsigned int)secs);
+    }
+
     if (heart_port) {
 	/* hearts input fd
 	 * We "know" drv_data is the in_fd since the port is started with read|write
 	 */
 	heart_fd[0] = (int)heart_port->drv_data;
 	heart_fd[1] = (int)driver_data[heart_fd[0]].ofd;
+	has_heart   = 1;
 
 	list = CONS(hp, make_small(8), list); hp += 2;
 
@@ -752,20 +768,14 @@ prepare_crash_dump(int secs)
 	erts_silence_warn_unused_result(nice(nice_val));
     }
 
-    /* Positive secs means an alarm must be set
-     * 0 or negative means no alarm
-     */
-    if (secs > 0) {
-	alarm((unsigned int)secs);
-    }
     UnUseTmpHeapNoproc(NUFBUF);
 #undef NUFBUF
+    return has_heart;
 }
 
-void
-erts_sys_prepare_crash_dump(int secs)
+int erts_sys_prepare_crash_dump(int secs)
 {
-    prepare_crash_dump(secs);
+    return prepare_crash_dump(secs);
 }
 
 static ERTS_INLINE void
@@ -802,12 +812,22 @@ static RETSIGTYPE request_break(int signum)
 static ERTS_INLINE void
 sigusr1_exit(void)
 {
-   /* We do this at interrupt level, since the main reason for
-      wanting to generate a crash dump in this way is that the emulator
-      is hung somewhere, so it won't be able to poll any flag we set here.
-      */
+    char env[21]; /* enough to hold any 64-bit integer */
+    size_t envsz;
+    int i, secs = -1;
+
+    /* We do this at interrupt level, since the main reason for
+     * wanting to generate a crash dump in this way is that the emulator
+     * is hung somewhere, so it won't be able to poll any flag we set here.
+     */
     ERTS_SET_GOT_SIGUSR1;
-    prepare_crash_dump((int)0);
+
+    envsz = sizeof(env);
+    if ((i = erts_sys_getenv_raw("ERL_CRASH_DUMP_SECONDS", env, &envsz)) >= 0) {
+	secs = i != 0 ? 0 : atoi(env);
+    }
+
+    prepare_crash_dump(secs);
     erl_exit(1, "Received SIGUSR1\n");
 }
 
diff --git a/erts/emulator/sys/vxworks/sys.c b/erts/emulator/sys/vxworks/sys.c
index 739b026fb1..3bdff5d7b6 100644
--- a/erts/emulator/sys/vxworks/sys.c
+++ b/erts/emulator/sys/vxworks/sys.c
@@ -296,10 +296,10 @@ void sys_sigrelease(int sig)
   sigprocmask(SIG_UNBLOCK, &mask, (sigset_t *)NULL);
 }
 
-void
+int
 erts_sys_prepare_crash_dump(void)
 {
-    
+    return 0;
 }
 
 /* register signal handlers XXX - they don't work, need to find out why... */
diff --git a/erts/emulator/sys/win32/sys.c b/erts/emulator/sys/win32/sys.c
index 50f6219374..578536ed08 100755
--- a/erts/emulator/sys/win32/sys.c
+++ b/erts/emulator/sys/win32/sys.c
@@ -255,8 +255,7 @@ void erl_sys_args(int* argc, char** argv)
 #endif
 }
 
-void
-erts_sys_prepare_crash_dump(int secs)
+int erts_sys_prepare_crash_dump(int secs)
 {
     Port *heart_port;
     Eterm heap[3];
@@ -271,10 +270,14 @@ erts_sys_prepare_crash_dump(int secs)
 
 	/* send to heart port, CMD = 8, i.e. prepare crash dump =o */
 	erts_write_to_port(NIL, heart_port, list);
+
+	return 1;
     }
 
     /* Windows - free file descriptors are hopefully available */
     /* Alarm not used on windows */
+
+    return 0;
 }
 
 static void