Bug 667562 – [0.11] Race condition between streaming thread and resetting pads in downwards state change

After an evaluation, GNOME has moved from Bugzilla to GitLab. Learn more about GitLab.
No new issues can be reported in GNOME Bugzilla anymore.
To report an issue in a GNOME project, go to GNOME GitLab.
Do not go to GNOME Gitlab for: Bluefish, Doxygen, GnuCash, GStreamer, java-gnome, LDTP, NetworkManager, Tomboy.

Bug 667562 - [0.11] Race condition between streaming thread and resetting pads in downwards state change


Summary:	[0.11] Race condition between streaming thread and resetting pads in downward...


Status:	RESOLVED FIXED

Product:	GStreamer
Classification:	Platform
Component:	gst-plugins-base
Version:	0.11.x
Hardware:	Other Linux

Importance:	Normal blocker
Target Milestone:	0.11.x
Assigned To:	GStreamer Maintainers
QA Contact:	GStreamer Maintainers

URL:
Whiteboard:

Depends on:
Blocks:

Reported:	2012-01-09 14:40 UTC by Sebastian Dröge (slomo)
Modified:	2012-07-19 10:13 UTC

See Also:
GNOME target:	---
GNOME version:	---

Attachments
audiodecoder: Don't assert on pad caps not being set (1.24 KB, patch) 2012-07-19 08:55 UTC, Edward Hervey	none	Details \| Review

Description Sebastian Dröge (slomo) 2012-01-09 14:40:07 UTC

Happens when stopping playback by setting state to READY, for example with the audio decoder base class:


** CRITICAL **: gst_audio_decoder_finish_frame: assertion `buf == NULL || gst_pad_has_current_caps (dec->srcpad)' failed
aborting...



(gdb) bt

+ Trace 229416

Thread 1 (Thread 0x7ffff7f3a920 (LWP 7032))

#0 __lll_lock_wait
at ../nptl/sysdeps/unix/sysv/linux/x86_64/lowlevellock.S line 136
#1 _L_lock_926
from /lib/x86_64-linux-gnu/libpthread.so.0
#2 __pthread_mutex_lock
at pthread_mutex_lock.c line 61
#3 g_static_rec_mutex_lock
at /tmp/buildd/glib2.0-2.28.6/./glib/gthread.c line 1424
#4 post_activate
at gstpad.c line 831
#5 gst_pad_activate_mode
at gstpad.c line 1014
#6 gst_pad_set_active
at gstpad.c line 893
#7 activate_pads
at gstelement.c line 2575
#8 gst_iterator_fold
at gstiterator.c line 614
#9 iterator_activate_fold_with_resync
at gstelement.c line 2595
#10 gst_element_pads_activate
at gstelement.c line 2639
#11 gst_element_change_state_func
at gstelement.c line 2703
#12 gst_audio_decoder_change_state
at gstaudiodecoder.c line 2075
#13 gst_element_change_state
at gstelement.c line 2490
#14 gst_element_set_state_func
at gstelement.c line 2446
#15 gst_bin_element_set_state
at gstbin.c line 2249
#16 gst_bin_change_state_func
at gstbin.c line 2554
#17 gst_decode_bin_change_state
at gstdecodebin2.c line 4041
#18 gst_element_change_state
at gstelement.c line 2490
#19 gst_element_set_state_func
at gstelement.c line 2446
#20 gst_bin_element_set_state
at gstbin.c line 2249
#21 gst_bin_change_state_func
at gstbin.c line 2554
#22 gst_uri_decode_bin_change_state
at gsturidecodebin.c line 2524
#23 gst_element_change_state
at gstelement.c line 2490
#24 gst_element_set_state_func
at gstelement.c line 2446
#25 gst_bin_element_set_state
at gstbin.c line 2249
#26 gst_bin_change_state_func
at gstbin.c line 2554
#27 gst_pipeline_change_state
at gstpipeline.c line 478
#28 gst_play_bin_change_state
at gstplaybin2.c line 3945
#29 gst_element_change_state
at gstelement.c line 2490
#30 gst_element_change_state
at gstelement.c line 2527
#31 gst_element_set_state_func
at gstelement.c line 2446
#32 stop_cb
at seek.c line 696
#33 stop_cb
at seek.c line 687
#34 delete_event_cb
at seek.c line 1772
#35 ??
from /usr/lib/libgtk-3.so.0
#36 g_closure_invoke
#37 signal_emit_unlocked_R
at /tmp/buildd/glib2.0-2.28.6/./gobject/gsignal.c line 3252
#38 g_signal_emit_valist
at /tmp/buildd/glib2.0-2.28.6/./gobject/gsignal.c line 2993
#39 g_signal_emit
at /tmp/buildd/glib2.0-2.28.6/./gobject/gsignal.c line 3040
#40 ??
from /usr/lib/libgtk-3.so.0
#41 gtk_main_do_event
from /usr/lib/libgtk-3.so.0
#42 ??
from /usr/lib/libgdk-3.so.0
#43 g_main_dispatch
at /tmp/buildd/glib2.0-2.28.6/./glib/gmain.c line 2440
#44 g_main_context_dispatch
at /tmp/buildd/glib2.0-2.28.6/./glib/gmain.c line 3013
#45 g_main_context_iterate
#46 g_main_loop_run
at /tmp/buildd/glib2.0-2.28.6/./glib/gmain.c line 3299
#47 gtk_main
from /usr/lib/libgtk-3.so.0
#48 main
at seek.c line 2152

Comment 1 Tim-Philipp Müller 2012-05-28 20:00:47 UTC

There were state-change related fixes lately, do those help with this by any chance?

Comment 2 Edward Hervey 2012-07-12 08:22:21 UTC

The behaviour it encounters is not fatal imho.

The pads being deactivated while the streaming thread is active is a common behaviour, and the expected result is to return GST_FLOW_WRONG_STATE.

Either we can use GST_PAD_IS_ACTIVE (srcpad) or we can store a local variable to know if we are active or not. And if it's not the case, return GST_FLOW_WRONG_STATE.

Comment 3 Edward Hervey 2012-07-19 08:55:54 UTC

Created attachment 219202 [details] [review]
audiodecoder: Don't assert on pad caps not being set

The decoder might have been de-activated in the meantime (resulting
in NULL pad caps).

If the decoder really isn't configured, then it will error out further
down when checking whether the GST_AUDIO_INFO_IS_VALID()

Comment 4 Edward Hervey 2012-07-19 10:13:48 UTC

commit 55f692eff682b78c924c5d79087e9f91e4dc93bb
Author: Edward Hervey <edward.hervey@collabora.co.uk>
Date:   Thu Jul 19 10:54:07 2012 +0200

    audiodecoder: Don't assert on pad caps not being set
    
    The decoder might have been de-activated in the meantime (resulting
    in NULL pad caps).
    
    If the decoder really isn't configured, then it will error out further
    down when checking whether the GST_AUDIO_INFO_IS_VALID()
    
    https://bugzilla.gnome.org/show_bug.cgi?id=667562