Bug 179356 – [GStreamer] Failure in gst_gl_context_thread_add

• Component: Media
• Status: RESOLVED DUPLICATE

[Michael Catanzaro]

There's another thread inside _mesa_glsl_compile_shader at the same time. Maybe that's bad?

Thread 1 (Thread 0x2b8428401700 (LWP 5903)):
#0  0x00002b83a02d292f in exec_node_remove () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/Mesa/src/glsl/list.h:170
#1  _ZN9exec_node6removeEv () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/Mesa/src/glsl/list.h:248
#2  _Z31optimize_dead_builtin_variablesP9exec_list16ir_variable_mode () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/Mesa/src/glsl/opt_dead_builtin_variables.cpp:79
#3  0x00002b83a0293cfb in _mesa_glsl_compile_shader () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/Mesa/src/glsl/glsl_parser_extras.cpp:1647
#4  0x00002b83a015fbba in compile_shader () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/Mesa/src/mesa/main/shaderapi.c:980
#5  0x00002b831fceb3d2 in _compile_shader () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/gst-plugins-bad-1.10.5/gst-libs/gst/gl/gstglslstage.c:486
#6  0x00002b831fcda990 in gst_gl_context_thread_add () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/gst-plugins-bad-1.10.5/gst-libs/gst/gl/gstglcontext.c:1537
#7  0x00002b831fcebe07 in gst_glsl_stage_compile () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/gst-plugins-bad-1.10.5/gst-libs/gst/gl/gstglslstage.c:542
#8  0x00002b831fce81eb in gst_gl_shader_compile_attach_stage () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/gst-plugins-bad-1.10.5/gst-libs/gst/gl/gstglshader.c:603
#9  0x00002b831fcee03f in _create_shader () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/gst-plugins-bad-1.10.5/gst-libs/gst/gl/gstglcolorconvert.c:1830
#10 _init_convert () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/gst-plugins-bad-1.10.5/gst-libs/gst/gl/gstglcolorconvert.c:2028
#11 _do_convert () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/gst-plugins-bad-1.10.5/gst-libs/gst/gl/gstglcolorconvert.c:2368
#12 0x00002b831fcf2683 in _run_message_sync () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/gst-plugins-bad-1.10.5/gst-libs/gst/gl/gstglwindow.c:601
#13 0x00002b831fcf2622 in _run_message_async () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/gst-plugins-bad-1.10.5/gst-libs/gst/gl/gstglwindow.c:668
#14 0x00002b8320cd15ca in g_main_dispatch () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/glib-2.52.1/glib/gmain.c:3212
#15 g_main_context_dispatch () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/glib-2.52.1/glib/gmain.c:3865
#16 0x00002b8320cd1948 in g_main_context_iterate () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/glib-2.52.1/glib/gmain.c:3938
#17 0x00002b8320cd1c62 in g_main_loop_run () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/glib-2.52.1/glib/gmain.c:4134
#18 0x00002b831fcf26f5 in gst_gl_window_default_run () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/gst-plugins-bad-1.10.5/gst-libs/gst/gl/gstglwindow.c:527
#19 0x00002b831fcdb95c in gst_gl_context_create_thread () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/gst-plugins-bad-1.10.5/gst-libs/gst/gl/gstglcontext.c:1273
#20 0x00002b8320cf8315 in g_thread_proxy () at /home/slave/webkitgtk/gtk-linux-64-release-tests/build/WebKitBuild/DependenciesGTK/Source/glib-2.52.1/glib/gthread.c:784
#21 0x00002b8323f2b494 in start_thread (arg=0x2b8428401700) at pthread_create.c:333
#22 0x00002b832510793f in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:97

Full backtrace here: https://build.webkit.org/results/GTK%20Linux%2064-bit%20Release%20(Tests)/r224517%20(4057)/imported/w3c/web-platform-tests/mediacapture-fromelement/capture-crash-log.txt

It's crashing imported/w3c/web-platform-tests/mediacapture-fromelement/capture.html so I'll update the expectations for that accordingly.

[Michael Catanzaro]

Note: it's flaky.

[Philippe Normand]

I can't make it fail here. Perhaps it's no longer flaky after the update to gst 1.12?

[Michael Catanzaro]

Last crash on the GTK release bot was r228208:

https://build.webkit.org/results/GTK%20Linux%2064-bit%20Release%20(Tests)/r228208%20(5378)/imported/w3c/web-platform-tests/mediacapture-fromelement/capture-crash-log.txt

so no, I don't think so.

[Michael Catanzaro]

I believe lots of tests were failing with this crash, BTW, so many that I wondered if we should undo the switch to GStreamerGL. I wonder if we have other bug reports for this.

[Philippe Normand]

Was this ever reported to Mesa and/or GStreamer bugzilla?

[Michael Catanzaro]

(In reply to Philippe Normand from comment #5)
> Was this ever reported to Mesa and/or GStreamer bugzilla?

Nope.

[Xabier Rodríguez Calvar]

I don't see a crash, just a failure.

[Philippe Normand]

Yeah well this specific trace is 3 years old, we did several Mesa and GStreamer updates since :)

Maybe update the TestExpectations then?

[Xabier Rodríguez Calvar]

(In reply to Philippe Normand from comment #8)
> Yeah well this specific trace is 3 years old, we did several Mesa and
> GStreamer updates since :)
> 
> Maybe update the TestExpectations then?

It does not appear as crash, just as failure, so I'll update the bug summary

[Philippe Normand]

Test is already marked as failing in the generic TestExpectations so no need to have another one in our ports.

Also since we don't support video.captureStream() anyway, I don't see how this test could pass, hence duplicate of bug 224972.

*** This bug has been marked as a duplicate of bug 224972 ***