Chromium needs to be able to import and display YV12 buffers coming from ARC++.
Additionally, being able to allocate and map YV12 buffer could be useful for media (libvpx could decode directly into a YV12 buffer).
Hi, I have a question about this test.
It is failing on Linux Desktop with GN (it does not to be include in GYP at all), e.g.
https://build.chromium.org/p/tryserver.chromium.linux/builders/linux_chromium_chromeos_ozone_rel_ng/builds/184706/steps/gl_unittests%20%28with%20patch%29%20on%20Ubuntu-12.04
Normally we don't run unit tests on target hardware. Do we run gl_unittests somewhere other than Linux Desktop (with chromeos=1) ?
The failure we are seeing is a crash:
[ RUN ] GLImageOzoneNativePixmapDrm/GLImageBindTest/0.BindTexImage
[32004:32004:0613/172359:543448650645:FATAL:gl_image_ozone_native_pixmap_drm_unittest.cc(50)] Check failed: ret.
#0 0x00000093389e base::debug::StackTrace::StackTrace()
#1 0x00000093d42a logging::LogMessage::~LogMessage()
#2 0x00000043c113 gl::(anonymous namespace)::GLImageOzoneNativePixmapDrmTestDelegate::CreateSolidColorImage()
#3 0x00000043cd71 gl::gtest_case_GLImageBindTest_::BindTexImage<>::TestBody()
#4 0x000000988a67 testing::Test::Run()
That test depends explicitly on drm and should not run on that bot. My intention was to build it only on a real cros device for now. Let me fix that.
The plan is to run those in the cros lab eventually.
OK. I'm pretty sure we don't do that currently (run gl_unittests on a target device). Mostly unit tests do not run on target devices because they are not dependent on what platform they are run on, but gl_unittests does seem like it would be an exception.
I think that what you want to test for is target_os == "chromeos", e.g.:
if (target_os == "chromeos" && target_cpu != "arm") {
Instead of
if (is_chromeos && target_cpu != "arm") {
Since is_chromeos refers to the UI and is true on linux chromeos.
If we can fix that quickly that would help get the linux chromeos ozone builder converted to GN sooner so we will catch these in the CQ, thanks!
Apologies, comment #7 provided misinformation. We can't actually easily distinguish whether we are building for linux-chromeos vs. for a device. We do run time tests for that (e.g. base::SysInfo::IsRunningOnChromeOS()).
I still think the change in Comment #13 is an improvement, but we will need to actually disable those tests then modify them to test for a valid environment.
You started fixing this bug over two years ago. Are you still working on it? You can update the status to "archived", "wontfix", or "closed". You can remove yourself as owner and change status to "untriaged", but if this is still a real bug, please do not sit on it.
Comment 1 by bugdroid1@chromium.org
, Jun 9 2016