Issue metadata
Sign in to add a comment
|
smoothness.scrolling_tough_ad_cases[.reference] flaky on Android One Perf (1) |
||||||||||||||||||||||
Issue descriptionRevision range first seen: as far as the bot's history goes (oldest: r398239:r398245 Link to failing step log: https://build.chromium.org/p/chromium.perf/builders/Android%20One%20Perf%20%281%29/builds/3092/steps/smoothness.scrolling_tough_ad_cases/logs/stdio In all cases, the failure is due to a native crash on http://www.forbes.com/sites/parmyolson/2015/07/29/jana-mobile-data-facebook-internet-org/. Most of them seem to be SVG related: (https://build.chromium.org/p/chromium.perf/builders/Android%20One%20Perf%20%281%29/builds/3056/steps/smoothness.scrolling_tough_ad_cases/logs/stdio) pid: 21911, tid: 21911, name: chromium.chrome >>> org.chromium.chrome <<< signal 6 (SIGABRT), code -6 (SI_TKILL), fault addr -------- r0 00000000 r1 00005597 r2 00000006 r3 00000000 r4 b6fbfe38 r5 00000006 r6 0000000b r7 0000010c r8 00000000 r9 00000000 sl 00000000 fp 000045f7 ip 00005597 sp beed3c10 lr b6e4cdb5 pc b6e73980 Stack Trace: RELADDR FUNCTION FILE:LINE 0003d980 tgkill+12 /system/lib/libc.so 00016db1 pthread_kill+52 /system/lib/libc.so 000179c7 raise+10 /system/lib/libc.so 00014169 __libc_android_abort+36 /system/lib/libc.so 000124f0 abort+4 /system/lib/libc.so 0138cc29 blink::SVGFEGaussianBlurElement::SVGFEGaussianBlurElement(blink::Document&) /b/build/slave/Android_Builder/build/src/third_party/WebKit/Source/core/svg/SVGFEGaussianBlurElement.cpp:33 (https://build.chromium.org/p/chromium.perf/builders/Android%20One%20Perf%20%281%29/builds/3016/steps/smoothness.scrolling_tough_ad_cases/logs/stdio pid: 10953, tid: 10953, name: chromium.chrome >>> org.chromium.chrome <<< signal 6 (SIGABRT), code -6 (SI_TKILL), fault addr -------- r0 00000000 r1 00002ac9 r2 00000006 r3 00000000 r4 b6f64e38 r5 00000006 r6 0000000b r7 0000010c r8 00000000 r9 00000000 sl 00000000 fp 000045f7 ip 00002ac9 sp be842c10 lr b6df1db5 pc b6e18980 Stack Trace: RELADDR FUNCTION FILE:LINE 0003d980 tgkill+12 /system/lib/libc.so 00016db1 pthread_kill+52 /system/lib/libc.so 000179c7 raise+10 /system/lib/libc.so 00014169 __libc_android_abort+36 /system/lib/libc.so 000124f0 abort+4 /system/lib/libc.so v------> blink::NormalPageArena::allocateObject(unsigned int, unsigned int) /b/build/slave/Android_Builder/build/src/third_party/WebKit/Source/platform/heap/HeapPage.h:912 v------> blink::ThreadHeap::allocateOnArenaIndex(blink::ThreadState*, unsigned int, int, unsigned int, char const*) /b/build/slave/Android_Builder/build/src/third_party/WebKit/Source/platform/heap/Heap.h:560 v------> blink::Node::allocateObject(unsigned int, bool) /b/build/slave/Android_Builder/build/src/third_party/WebKit/Source/core/dom/Node.h:180 v------> blink::Node::operator new(unsigned int) /b/build/slave/Android_Builder/build/src/third_party/WebKit/Source/core/dom/Node.h:174 0136fdc1 blink::SVGFEFuncGElement::create(blink::Document&) /b/build/slave/Android_Builder/build/src/third_party/WebKit/Source/core/svg/SVGFEFuncGElement.cpp:30 (https://build.chromium.org/p/chromium.perf/builders/Android%20One%20Perf%20%281%29/builds/3084/steps/smoothness.scrolling_tough_ad_cases/logs/stdio) pid: 18679, tid: 18679, name: chromium.chrome >>> org.chromium.chrome <<< signal 6 (SIGABRT), code -6 (SI_TKILL), fault addr -------- r0 00000000 r1 000048f7 r2 00000006 r3 00000000 r4 b6f69e38 r5 00000006 r6 0000000b r7 0000010c r8 00000000 r9 00000000 sl 00000000 fp 000045f7 ip 000048f7 sp bea69c10 lr b6df6db5 pc b6e1d980 Stack Trace: RELADDR FUNCTION FILE:LINE 0003d980 tgkill+12 /system/lib/libc.so 00016db1 pthread_kill+52 /system/lib/libc.so 000179c7 raise+10 /system/lib/libc.so 00014169 __libc_android_abort+36 /system/lib/libc.so 000124f0 abort+4 /system/lib/libc.so v------> blink::Member<blink::SVGAnimatedNumber>::Member(blink::SVGAnimatedNumber*) /mnt/data/b/build/slave/Android_Builder/build/src/third_party/WebKit/Source/platform/heap/Member.h:32 01387c79 blink::SVGFELightElement::SVGFELightElement(blink::QualifiedName const&, blink::Document&) /mnt/data/b/build/slave/Android_Builder/build/src/third_party/WebKit/Source/core/svg/SVGFELightElement.cpp:43 and one of them was due to a failed check in gpu::gles2 (possibly unrelated as it was in the reference benchmark): (https://build.chromium.org/p/chromium.perf/builders/Android%20One%20Perf%20%281%29/builds/3081/steps/smoothness.scrolling_tough_ad_cases.reference/logs/stdio) pid: 12564, tid: 12564, name: oid.apps.chrome >>> com.google.android.apps.chrome <<< signal 6 (SIGABRT), code -6 (SI_TKILL), fault addr -------- [FATAL:synthetic_gesture_target_base.cc(62)] Check failed: web_touch.touches[i].state != WebTouchPoint::StatePressed || PointIsWithinContents(web_touch.touches[i].position.x, web_touch.touches[i].position.y). Touch coordinates are not within content bounds on TouchStart. r0 00000000 r1 00003114 r2 00000006 r3 00000000 r4 b6f78e38 r5 00000006 r6 0000000b r7 0000010c r8 b6e4add4 r9 bee69dd0 sl 00000000 fp 9ab35f28 ip 00003114 sp bee698a0 lr b6e05db5 pc b6e2c980 Stack Trace: RELADDR FUNCTION FILE:LINE 0003d980 tgkill+12 /system/lib/libc.so 00016db1 pthread_kill+52 /system/lib/libc.so 000179c7 raise+10 /system/lib/libc.so 00014169 __libc_android_abort+36 /system/lib/libc.so 000124f0 abort+4 /system/lib/libc.so 004d00bd gpu::gles2::Program::ClearUniforms(std::__1::vector<unsigned char, std::__1::allocator<unsigned char> >*) /mnt/data/b/build/slave/Android_Builder/build/src/gpu/command_buffer/service/program_manager.cc:430 If the test is disabled, please downgrade to Pri-2.
,
Jun 21 2016
This strongly suggests running out of memory due to a SVG filter tree. There are no SVG changes in the reported revision range on the bot. In the bots reported revision range, this is the only change that seems to materially affect memory. There are refactoring changes that, if made incorrectly, might also affect memory. https://chromium.googlesource.com/chromium/src/+/0ca25bf2ec49526c331b85a7e29d4c4f5aa2db75 I vote for disabling the test. Can we do that only for Android One?
,
Jun 21 2016
,
Jun 21 2016
,
Jun 21 2016
Any chance there are more stack context available? (Seeing SVGFELightElement feels a bit odd...)
,
Jun 22 2016
fs: Please find attached the stack contexts of all threads for the 4 failures in the bug description. I don't have any more details unfortunately.
,
Jun 22 2016
Thanks! Not much more that can be read out of that though - lot's of filter related code in various threads (FEImage et.c) and some SVGImage stuff (SVGImageChromeClient::create appearing multiple times in stacks_2.) Makes one suspect SVG images with filters might be appearing en masse in the content - maybe the same thing loaded multiple times even.
,
Jun 23 2016
#2: Is disabling the benchmark the right thing to do? Isn't this an actual memory regression that needs to be looked into by someone working on SVG? I'm re-assigning this to test owner to decide what to do.
,
Jul 18 2016
,
Aug 2 2016
Firendly sheriff ping: skyostil@ Could you please respond on #8
,
Aug 4 2016
tough_ad_cases is a collection of real web sites, so it's pretty likely find memory issues like these. My vote would be to find someone to investigate why this is crashing and disable in meanwhile to keep the bot green.
,
Aug 29 2016
sheriff ping!
,
Oct 3 2016
This is fixed, right? It is not disabled on Android One now. Marking as fixed. Re-open if needed. |
|||||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||||
Comment 1 by petrcermak@chromium.org
, Jun 21 2016Labels: OS-Android