Flaky crashes in WebGL2 tests, probably caused by V8 |
||||
Issue descriptionhttps://ci.chromium.org/p/chromium/builders/luci.chromium.ci/Linux%20FYI%20Release%20%28Intel%20HD%20630%29/6773 webgl2_conformance_tests WebglConformance_conformance_ogles_GL_mat_mat_017_to_024 https://ci.chromium.org/p/chromium/builders/luci.chromium.ci/Linux%20FYI%20Release%20%28Intel%20HD%20630%29/6774 webgl2_conformance_tests WebglConformance_deqp_functional_gles3_shaderoperator_common_functions_02 https://ci.chromium.org/p/chromium/builders/luci.chromium.ci/Linux%20FYI%20Release%20%28Intel%20HD%20630%29/6775 webgl2_conformance_gl_passthrough_tests WebglConformance_deqp_functional_gles3_fbocolorbuffer_tex3d_04 WebglConformance_deqp_functional_gles3_texturefiltering_cube_sizes_02 https://ci.chromium.org/p/chromium/builders/luci.chromium.ci/Linux%20FYI%20Release%20%28NVIDIA%29/5990 webgl2_conformance_gl_passthrough_tests WebglConformance_deqp_functional_gles3_framebufferblit_depth_stencil https://ci.chromium.org/p/chromium/builders/luci.chromium.ci/Linux%20FYI%20Release%20%28NVIDIA%29/5991 webgl2_conformance_gl_passthrough_tests WebglConformance_deqp_functional_gles3_shaderindexing_mat_01 Crash is: 0 chrome!Lookup<WTF::IdentityHashTranslator<WTF::MemberHash<blink::Animation>, WTF::HashMapValueTraits<WTF::HashTraits<blink::WeakMember<blink::Animation> >, WTF::HashTraits<unsigned int> >, blink::HeapAllocator>, blink::Animation *> [member.h : 531 + 0x0] 1 chrome!RemoveClient [hash_table.h : 1520 + 0x8] 2 chrome!SetResource [resource_client.cc : 44 + 0x8] 3 chrome!DisposeInternal [resource_client.h : 75 + 0x9] 4 chrome!Dispose [pending_script.cc : 76 + 0x9] 5 chrome!ExecuteScriptBlock [pending_script.cc : 181 + 0x8] 6 chrome!ExecuteAsyncTask [script_runner.cc : 242 + 0xc] 7 chrome!blink::ScriptRunner::ExecuteTask() + 0x17 8 chrome!Run [callback.h : 99 + 0x3] 9 chrome!RunTask [callback.h : 99 + 0x3] 10 chrome!DoWork [thread_controller_impl.cc : 196 + 0x1a] 11 chrome!Run [bind_internal.h : 516 + 0x3] 12 chrome!RunTask [callback.h : 99 + 0x3] 13 chrome!RunTask [message_loop.cc : 550 + 0xf] 14 chrome!DoWork [message_loop.cc : 561 + 0x12] 15 chrome!Run [message_pump_default.cc : 37 + 0x9] 16 chrome!Run [message_loop.cc : 502 + 0x9] 17 chrome!Run [run_loop.cc : 102 + 0x9] 18 chrome!RendererMain [renderer_main.cc : 202 + 0xc] 19 chrome!RunZygote [content_main_runner_impl.cc : 495 + 0x4] 20 chrome!Run [content_main_runner_impl.cc : 906 + 0x5] 21 chrome!Main [main.cc : 472 + 0xa] 22 chrome!content::ContentMain(content::ContentMainParams const&) + 0x51 23 chrome!ChromeMain [chrome_main.cc : 102 + 0x5] 24 libc-2.19.so + 0x21f45 25 chrome!frame_dummy + 0x30 26 chrome + 0x2d1f000 27 ld-2.19.so + 0x101f3 28 chrome + 0x2d1f000 29 chrome!_start + 0x2a 30 0x7ffcef1c9968 https://ci.chromium.org/p/chromium/builders/luci.chromium.ci/Linux%20FYI%20Release%20%28NVIDIA%29/5992 webgl2_conformance_tests WebglConformance_deqp_functional_gles3_textureformat_sized_color_cube_pot_01 This one has a different stack: 0 chrome!Visit [heap_page.h : 906 + 0x0] 1 chrome!Trace<blink::Visitor *> [visitor.h : 113 + 0x20] 2 chrome!Visit [marking_visitor.h : 103 + 0xa] 3 chrome!Trace [visitor.h : 123 + 0x17] 4 chrome!Visit [marking_visitor.h : 103 + 0xa] 5 chrome!Trace<blink::Visitor *> [visitor.h : 113 + 0x20] 6 chrome!Visit [marking_visitor.h : 103 + 0xa] 7 chrome!Trace [visitor.h : 123 + 0x17] 8 chrome!AdvanceMarking [heap.cc : 260 + 0x9] 9 chrome!MarkPhaseAdvanceMarking [thread_state.cc : 1781 + 0x8] 10 chrome!IncrementalMarkingStep [thread_state.cc : 1500 + 0xb] 11 chrome!RunScheduledGC [thread_state.cc : 963 + 0x8] 12 chrome!SafePoint [thread_state.cc : 1314 + 0xb] 13 chrome!NotifyDidProcessTask [sequence_manager_impl.cc : 573 + 0xc] 14 chrome!DidRunTask [sequence_manager_impl.cc : 421 + 0xf] 15 chrome!DoWork [thread_controller_impl.cc : 202 + 0x6] 16 chrome!Run [bind_internal.h : 516 + 0x3] 17 chrome!RunTask [callback.h : 99 + 0x3] 18 chrome!RunTask [message_loop.cc : 550 + 0xf] 19 chrome!DoWork [message_loop.cc : 561 + 0x12] 20 chrome!Run [message_pump_default.cc : 37 + 0x9] 21 chrome!Run [message_loop.cc : 502 + 0x9] 22 chrome!Run [run_loop.cc : 102 + 0x9] 23 chrome!RendererMain [renderer_main.cc : 202 + 0xc] 24 chrome!RunZygote [content_main_runner_impl.cc : 495 + 0x4] 25 chrome!Run [content_main_runner_impl.cc : 906 + 0x5] 26 chrome!Main [main.cc : 472 + 0xa] 27 chrome!content::ContentMain(content::ContentMainParams const&) + 0x51 28 chrome!ChromeMain [chrome_main.cc : 102 + 0x5] 29 libc-2.19.so + 0x21f45 30 chrome!frame_dummy + 0x30 31 chrome + 0x2d1f000 32 ld-2.19.so + 0x101f3 33 chrome + 0x2d1f000 34 chrome!_start + 0x2a 35 0x7ffee6f1a4a8 https://ci.chromium.org/p/chromium/builders/luci.chromium.ci/Linux%20FYI%20Release%20%28Intel%20HD%20630%29/6773 gives regression range crrev.com/603467..603477 https://ci.chromium.org/p/chromium/builders/luci.chromium.ci/Linux%20FYI%20Release%20(NVIDIA) gives regression range crrev.com/603474..603481 Intersection is crrev.com/603474..603477 Most likely culprits are: V8 roll in https://chromium-review.googlesource.com/c/chromium/src/+/1304402 Enable HeapCompaction in https://chromium-review.googlesource.com/c/chromium/src/+/1297782 This affects Chromium CQ, see https://ci.chromium.org/p/chromium/builders/luci.chromium.try/linux_optional_gpu_tests_rel/10899, ironically, for V8 roll. Could you please revert one or both of the CLs above?
,
Oct 29
The major change there that sticks out is enabling compaction: 10a10dd Oilpan: Enable HeapCompaction for incremental marking by Keishi Hattori ยท 5 hours ago Since this seems to be blocking rolls and whatnot I will speculatively revert for now. We can reland if it turns out that it was a different issue.
,
Oct 29
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/0ce72e129d923a9f4a6ff0eae3c599b25fdb8d36 commit 0ce72e129d923a9f4a6ff0eae3c599b25fdb8d36 Author: Yuly Novikov <ynovikov@chromium.org> Date: Mon Oct 29 18:33:30 2018 Revert "Oilpan: Enable HeapCompaction for incremental marking" This reverts commit 10a10dd9fde23fa0f038c83fa9940ea69afcae2a. Reason for revert: Suspected for flaky crashes affecting Chromium CQ. Original change's description: > Oilpan: Enable HeapCompaction for incremental marking > > Enables HeapCompaction for kIncrementalIdleGC and kIncrementalV8FollowupGC. > > Bug: 864425 > > Change-Id: I1fb04d7c501a8519e106237946f86f52eb7ce8f8 > Reviewed-on: https://chromium-review.googlesource.com/c/1297782 > Reviewed-by: Kentaro Hara <haraken@chromium.org> > Reviewed-by: Michael Lippautz <mlippautz@chromium.org> > Commit-Queue: Keishi Hattori <keishi@chromium.org> > Cr-Commit-Position: refs/heads/master@{#603477} TBR=haraken@chromium.org,keishi@chromium.org,mlippautz@chromium.org Change-Id: Ie72a2eb04571945b3515ee7374a75a1ab87275c2 No-Presubmit: true No-Tree-Checks: true No-Try: true Bug: 864425, 899809 Reviewed-on: https://chromium-review.googlesource.com/c/1305833 Reviewed-by: Yuly Novikov <ynovikov@chromium.org> Commit-Queue: Yuly Novikov <ynovikov@chromium.org> Cr-Commit-Position: refs/heads/master@{#603573} [modify] https://crrev.com/0ce72e129d923a9f4a6ff0eae3c599b25fdb8d36/third_party/blink/renderer/platform/heap/heap_compact.cc
,
Oct 29
Sorry, forgot to click "Save changes" on this: According to https://ci.chromium.org/p/chromium/g/chromium.gpu.fyi/console V8 sheriff is <none>, thus going to tentatively revert "Enable HeapCompaction".
,
Oct 29
Well, the bots are green since the revert landed: https://ci.chromium.org/p/chromium/builders/luci.chromium.ci/Linux%20FYI%20Release%20%28Intel%20HD%20630%29/6778 https://ci.chromium.org/p/chromium/builders/luci.chromium.ci/Linux%20FYI%20Release%20%28NVIDIA%29/5995 However, they became green in the build before the revert, so I'm not 100% certain. But, I don't see anything related in the range of the previous green builds, and the reverted CL is still most likely culprit in regression range. Leaving to keishi@ to investigate further.
,
Oct 29
,
Nov 5
I landed a fix. Hopefully it fixed this. https://chromium-review.googlesource.com/c/chromium/src/+/1312189 |
||||
►
Sign in to add a comment |
||||
Comment 1 by ynovikov@chromium.org
, Oct 29Summary: Flaky crashes in WebGL2 tests, probably caused by V8 (was: Flaky crashes in WebGL2 CTS on Linux Intel and NVIDIA)