New issue
Advanced search Search tips

Issue 899985 link

Starred by 2 users

Issue metadata

Status: Assigned
Owner:
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Mac
Pri: 1
Type: Bug



Sign in to add a comment

GrCacheControllerTest.ResetPurgeGrCacheOnReuse fails on first run, passes when retried in isolated

Project Member Reported by w...@chromium.org, Oct 29

Issue description

gpu_unittests:GrCacheControllerTest.ResetPurgeGrCacheOnReuse fails on Mac when run in a batch, only passing when retried in isolation:

[ RUN      ] GrCacheControllerTest.ResetPurgeGrCacheOnReuse
[11088:775:1027/125637.168658:10563824441610:FATAL:gl_bindings_autogen_egl.cc(28)] Check failed: this_bytes[0] == 0.
0   gpu_unittests                       0x000000010ec5f05f base::debug::StackTrace::StackTrace(unsigned long) + 31
1   gpu_unittests                       0x000000010eb73d9f logging::LogMessage::~LogMessage() + 223
2   gpu_unittests                       0x000000010f198dd6 gl::DriverEGL::InitializeStaticBindings() + 86
3   gpu_unittests                       0x000000010f1a37a5 gl::InitializeStaticGLBindingsEGL() + 21
4   gpu_unittests                       0x000000010f9a7feb gl::init::InitializeStaticGLBindings(gl::GLImplementation) + 859
5   gpu_unittests                       0x000000010f9a6b44 gl::init::InitializeGLOneOffImplementation(gl::GLImplementation, bool, bool, bool, bool) + 36
6   gpu_unittests                       0x000000010f3d0405 gl::(anonymous namespace)::InitializeOneOffHelper(bool) + 373
7   gpu_unittests                       0x000000010d2190da gpu::raster::GrCacheControllerTest::SetUp() + 26
8   gpu_unittests                       0x000000010d896bfd testing::Test::Run() + 109
9   gpu_unittests                       0x000000010d8978f0 testing::TestInfo::Run() + 288
10  gpu_unittests                       0x000000010d897f17 testing::TestCase::Run() + 247
11  gpu_unittests                       0x000000010d8a5157 testing::internal::UnitTestImpl::RunAllTests() + 855
12  gpu_unittests                       0x000000010d8a4ddd testing::UnitTest::Run() + 109
13  gpu_unittests                       0x000000010ec83c86 base::TestSuite::Run() + 166
14  gpu_unittests                       0x000000010ec93c4e base::(anonymous namespace)::LaunchUnitTestsInternal(base::OnceCallback<int ()>, unsigned long, int, bool, base::OnceCallback<void ()>) + 382
15  gpu_unittests                       0x000000010ec93aa1 base::LaunchUnitTests(int, char**, base::OnceCallback<int ()>) + 161
16  gpu_unittests                       0x000000010cde7d76 main + 166
17  libdyld.dylib                       0x00007fff7bd53015 start + 1
18  ???                                 0x0000000000000008 0x0 + 8

It looks like some preceding test is failing to invoke ClearBindingsEGL() before it exits (i.e. please don't just disable this test ;)
 
Cc: penghuang@chromium.org
Owner: khushals...@chromium.org
Status: Assigned (was: Untriaged)
khushal: Assigning to you as I think you're the author.  PTAL.
Cc: w...@chromium.org
Is there some bot this is failing on? The test has been up for a while, I'm curious if the failure started recently.
Re #2: It's quite possible that it has always failed on first run; the issue is that some other test is leaking the global EGL driver initialization (i.e. calling InitializeStaticBindings without later calling ClearBindingsEGL).

This wouldn't cause the bot to fail, since TestLauncher defaults to retrying each test up to three times.  I believe this was failing on a Mac GPU bot, but I'll have to check.

Not sure why it doesn't show in FindIt - perhaps FindIt doesn't watch that bot?
Hmmm. Let me try to repro this locally with a repeat-count of 1. I looked at all the tests in the gpu_unittests suite and seemed like all of those were shutting down GL state correctly.
Best way to repro is to run the suite with --test-launcher-retry-limit=0, or with --single-process-tests.

Sign in to add a comment