As part of bug 1403283 re-enabling the unit tests on Autophone, I've found the following failures preventing Mdm1 from being enabled. I ran two sets of tests using production and staging devices: <https://treeherder.mozilla.org/#/jobs?repo=try&revision=5213cf3519eb9fa4ed16db023a46229c3d49d1f2&filter-searchStr=Mdm1> <https://treeherder.allizom.org/#/jobs?repo=try&revision=5213cf3519eb9fa4ed16db023a46229c3d49d1f2&filter-searchStr=Mdm1> opt: TEST-UNEXPECTED-FAIL | dom/media/test/test_mediarecorder_record_addtracked_stream.html | Test timed out. debug: PROCESS-CRASH | dom/media/test/test_mediarecorder_record_addtracked_stream.html | application crashed [@ mozilla::gfx::Log<1, mozilla::gfx::CriticalLogger>::~Log] Crash dump filename: /tmp/tmpxtlSC9/717e8999-cd84-552d-8bcd-0f48cb31fb0f.dmp Operating system: Android 0.0.0 Linux 3.4.0-gadb2201 #1 SMP PREEMPT Wed Nov 20 14:42:53 PST 2013 armv7l CPU: arm ARMv7 Qualcomm Krait features: swp,half,thumb,fastmult,vfpv2,edsp,neon,vfpv3,tls,vfpv4,idiva,idivt 4 CPUs GPU: UNKNOWN Crash reason: SIGSEGV Crash address: 0x0 Process uptime: not available Thread 11 (crashed) 0 libxul.so!mozilla::gfx::Log<1, mozilla::gfx::CriticalLogger>::~Log [Logging.h:2d7b8b5dd174 : 521 + 0x10] r0 = 0x0000007e r1 = 0x8e155b27 r2 = 0x8e155b27 r3 = 0x00000000 r4 = 0x78f4c3a0 r5 = 0x75c14c48 r6 = 0x00000000 r7 = 0x75c14c88 r8 = 0x75c14e28 r9 = 0x75c14dfc r10 = 0x85550240 r12 = 0x00000003 fp = 0x75c14e14 sp = 0x75c14c48 lr = 0x76af836b pc = 0x76af9142 Found by: given as instruction pointer in context 1 libxul.so!mozilla::gl::GLBlitHelper::BlitImage [GLBlitHelper.cpp:2d7b8b5dd174 : 643 + 0x5] r4 = 0x75c14c7c r5 = 0x85550240 r6 = 0x00000000 r7 = 0x75c14f14 r8 = 0x75c14e28 r9 = 0x75c14dfc r10 = 0x85550240 fp = 0x75c14e14 sp = 0x75c14c78 pc = 0x76b6187b Found by: call frame info 2 libxul.so!mozilla::gl::GLBlitHelper::BlitImageToFramebuffer [GLBlitHelper.cpp:2d7b8b5dd174 : 610 + 0x3] r4 = 0x00000005 r5 = 0x85550240 r6 = 0x00000000 r7 = 0x75c14f14 r8 = 0x75c14e28 r9 = 0x75c14dfc r10 = 0x85550240 fp = 0x75c14e14 sp = 0x75c14d20 pc = 0x76b62f0f Found by: call frame info 3 libxul.so!mozilla::layers::GLImage::GetAsSourceSurface [GLImages.cpp:2d7b8b5dd174 : 54 + 0x9] r4 = 0x75c14e3c r5 = 0x7a370d34 r6 = 0x00000000 r7 = 0x75c14f14 r8 = 0x75c14e28 r9 = 0x75c14dfc r10 = 0x85550240 fp = 0x75c14e14 sp = 0x75c14dd0 pc = 0x76b7f1b7 Found by: call frame info 4 libxul.so!nsLayoutUtils::SurfaceFromElement [nsLayoutUtils.cpp:2d7b8b5dd174 : 7652 + 0x5] r4 = 0x75c15030 r5 = 0x75c14f14 r6 = 0x857b4800 r7 = 0x75c14f0c r8 = 0x00000000 r9 = 0x00000000 r10 = 0x857bb84c fp = 0x75c14fd8 sp = 0x75c14f00 pc = 0x77952035 Found by: call frame info 5 libxul.so!nsLayoutUtils::SurfaceFromElement [nsLayoutUtils.cpp:2d7b8b5dd174 : 7685 + 0xb] r4 = 0x75c15030 r5 = 0x85575580 r6 = 0x857b4800 r7 = 0x86b421f0 r8 = 0x857bb84c r9 = 0x00000012 r10 = 0x00000000 fp = 0x75c14fd8 sp = 0x75c14f40 pc = 0x77971145 Found by: call frame info 6 libxul.so!mozilla::dom::CanvasRenderingContext2D::DrawImage [CanvasRenderingContext2D.cpp:2d7b8b5dd174 : 5191 + 0x11] r4 = 0x857bb800 r5 = 0x75c14fc8 r6 = 0x75c14fbc r7 = 0x857b4800 r8 = 0x75c14fb0 r9 = 0x75c151b8 r10 = 0x00000000 fp = 0x75c14fd8 sp = 0x75c14f68 pc = 0x7723a791 Found by: call frame info 7 libxul.so!mozilla::dom::CanvasRenderingContext2DBinding::drawImage [CanvasRenderingContext2D.h:2d7b8b5dd174 : 241 + 0x4b] r4 = 0x75c15244 r5 = 0x00000001 r6 = 0x75e2f800 r7 = 0x00000000 r8 = 0x75c151c8 r9 = 0x857bb800 r10 = 0x00000000 fp = 0x7ff00000 sp = 0x75c15120 pc = 0x76fe7aeb Found by: call frame info test added in bug 1296531.

This looks like a graphics issue. Crashing at [1]. We're hitting this after recording a video with MediaRecorder, then trying to draw its frames to a 2d canvas. This can be tracked in the test where we're starting this drawing at [2]. The last log entry before the crash (except for two flaky timeout warnings) is: > {"action":"test_status","time":1507360440227,"thread":null,"pid":null,"source":"mochitest","test":"/tests/dom/media/test/test_mediarecorder_record_addtracked_stream.html","subtest":"Recording should have one audio track","status":"PASS","js_source":"TestRunner.js"} John, would this fall under Firefox for Android::Audio/Video (because of the decoder outputting these images) or Core::Graphics (for the conversion failing)? [1] http://searchfox.org/mozilla-central/rev/ed1d5223adcdc07e9a2589ee20f4e5ee91b4df10/gfx/gl/GLBlitHelper.cpp#643 [2] http://searchfox.org/mozilla-central/rev/ed1d5223adcdc07e9a2589ee20f4e5ee91b4df10/dom/media/test/test_mediarecorder_record_addtracked_stream.html#129

Bug 1322650 (patch [1]) looks like it might have caused a regression here. snorp, can you take a look? [1] http://searchfox.org/mozilla-central/diff/06131003608d8b7a1be322265ccc5362e2986515/gfx/gl/GLBlitHelper.cpp#685

(In reply to Andreas Pehrson [:pehrsons] from comment #1) > This looks like a graphics issue. Crashing at [1]. > > John, would this fall under Firefox for Android::Audio/Video (because of the > decoder outputting these images) or Core::Graphics (for the conversion > failing)? > > > [1] > http://searchfox.org/mozilla-central/rev/ > ed1d5223adcdc07e9a2589ee20f4e5ee91b4df10/gfx/gl/GLBlitHelper.cpp#643 > [2] > http://searchfox.org/mozilla-central/rev/ > ed1d5223adcdc07e9a2589ee20f4e5ee91b4df10/dom/media/test/ > test_mediarecorder_record_addtracked_stream.html#129 Core::Graphics, IMHO. On Android, SurfaceTexture was used to convert video decoder output buffers into GL textures for gfx to consume, and should work for both compositor and canvas.

Did a quick search in the codebase. It looks like GLReadTexImageHelper::ReadTexImage() [1] performs similar tasks as current GLImage::GetAsSourceSurface() implementation. Perhaps introduce a method (to TexturePoolOGL, which generated the texture and knows the required GL context) that sets up a helper and reads texture images data would simplify the code? [1] http://searchfox.org/mozilla-central/source/gfx/gl/GLReadTexImageHelper.cpp#501

We're fixing the SurfaceTexture blitting in bug 1395497.

Disable test_mediarecorder_record_addtracked_stream.html on Android.

Pushed by bclary@mozilla.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/8172eea4088e disable Mdm1 dom/media/test/test_mediarecorder_record_addtracked_stream, r=jmaher.

re-enable Mdm1.

https://github.com/mozilla/autophone/commit/44fb480968fbfd8ba81b0cd3d0ccf89c588f9b06 deploying shortly.