Changes between Version 2 and Version 3 of helgrind

Timestamp:

Jan 14, 2020 5:37:40 AM (4 years ago)

Author:

ming

Comment:

--

helgrind

@@ -106 +106 @@
 }}}
 
-3. Pulse Audio related[[br]]
-  Status: outside PJSIP's scope[[br]]
-  TODO: create a suppression file for this
-
-{{{
-==26641== Possible data race during read of size 4 at 0x6CEBC28 by thread #1
-==26641== Locks held: none
-==26641==    at 0x6AB5FE1: pa_once_begin (in
-/usr/lib/x86_64-linux-gnu/pulseaudio/libpulsecommon-4.0.so)
-==26641==    by 0x6AB616A: pa_run_once (in
-/usr/lib/x86_64-linux-gnu/pulseaudio/libpulsecommon-4.0.so)
-==26641==    by 0x6ACAE43: pa_thread_self (in
-/usr/lib/x86_64-linux-gnu/pulseaudio/libpulsecommon-4.0.so)
-==26641==    by 0x666146B: pa_threaded_mainloop_lock (in
-/usr/lib/x86_64-linux-gnu/libpulse.so.0.16.2)
-==26641==    by 0x8F573AF: pulse_connect (in
-/usr/lib/x86_64-linux-gnu/alsa-lib/libasound_module_pcm_pulse.so)
-==26641==    by 0x8F56C9B: _snd_pcm_pulse_open (in
-/usr/lib/x86_64-linux-gnu/alsa-lib/libasound_module_pcm_pulse.so)
-==26641==    by 0x55B36FC: ??? (in /usr/lib/x86_64-linux-gnu/libasound.so.2.0.0)
-==26641==    by 0x55B3CF5: ??? (in /usr/lib/x86_64-linux-gnu/libasound.so.2.0.0)
-==26641==    by 0x559982: OpenPcm (pa_linux_alsa.c:544)
-==26641==    by 0x559A34: FillInDevInfo (pa_linux_alsa.c:580)
-==26641==    by 0x55A9B8: BuildDeviceList (pa_linux_alsa.c:853)
-==26641==    by 0x558E8B: PaAlsa_Initialize (pa_linux_alsa.c:266)
-==26641==
-==26641== This conflicts with a previous write of size 4 by thread #2
-==26641== Locks held: 1, at address 0x5E41130
-==26641==    at 0x6AB6093: pa_once_end (in
-/usr/lib/x86_64-linux-gnu/pulseaudio/libpulsecommon-4.0.so)
-==26641==    by 0x6ACAEEC: ??? (in
-/usr/lib/x86_64-linux-gnu/pulseaudio/libpulsecommon-4.0.so)
-==26641==    by 0x4C30E26: mythread_wrapper (hg_intercepts.c:233)
-==26641==    by 0x5353181: start_thread (pthread_create.c:312)
-==26641==    by 0x595400C: clone (clone.S:111)
-}}}
-
-4. Data race over access to g_last_thread variable.
+3. Data race over access to g_last_thread variable.
 Status: won't fix. The variable is only used for information purpose only, while mutex locking for logging will be quite expensive.
 
@@ -200 +163 @@
 }}}
 
-5. Data race over cp->used_size variable in pool_caching.
+4. Data race over cp->used_size variable in pool_caching.
 Status: won't fix, the variable is for statistics purpose only.
 
@@ -241 +204 @@
 }}}
 
-6. Data race over tcp->pending_connect variable[[br]]
-Status: should be safe. A modification to suppress this would be desirable.
+5. Data race over tcp->has_pending_connect variable[[br]]
+Status: intentional. For optimisation, we intentionally check if has_pending_connect is true before holding the lock and checking the variable again.
 
 {{{
@@ -329 +292 @@
 }}}
 
-7. Destroying a locked mutex in sip dialog
+6. Destroying a locked mutex in sip dialog
 Status: won't fix. When calling pj_mutex_destroy() there is actually no requirement that the mutex should be unlocked. If the mutex destroy fails, pj_mutex_destroy() will automatically retry several times by unlocking the mutex first.
 {{{