Copied from issue 707711
Running this script below on Linux with release asan netted me two failures in LayerTreeHostTilesTestPartialInvalidation.PartialRaster_SingleThread_GpuRaster
--snip--
set -e
ninja -C out/asan cc_unittests -j500 -l20
while true; do
out/asan/cc_unittests --gtest_filter=*LayerTreeHostTilesTestPartialInvalidation.*
done
--snip--
Error #1 (pixels attached as images below)
[ RUN ] LayerTreeHostTilesTestPartialInvalidation.PartialRaster_SingleThread_GpuRaster
[21858:21924:0509/105051.573894:2408701337206:ERROR:gles2_cmd_decoder.cc(7783)] [.GpuRasterization]GL ERROR :GL_INVALID_OPERATION : glFramebufferTexture2D: unknown texture_ref
[21858:22222:0509/105051.575345:2408701338663:ERROR:gpu_raster_buffer_provider.cc(54)] Failed to allocate raster surface
[21858:21858:0509/105052.062227:2408701825546:ERROR:pixel_comparator.cc(47)] Number of pixel with an error: 30000
[21858:21858:0509/105052.062360:2408701825666:ERROR:pixel_comparator.cc(48)] Error Bounding Box : 0,0 200x200
[21858:21858:0509/105052.085300:2408701848619:ERROR:pixel_test_utils.cc(79)] Pixels do not match!
../../cc/test/layer_tree_pixel_test.cc:117: Failure
Value of: MatchesPNGFile(*result_bitmap_, ref_file_path, *pixel_comparator_)
Actual: false
Expected: true
[ FAILED ] LayerTreeHostTilesTestPartialInvalidation.PartialRaster_SingleThread_GpuRaster (597 ms)
Error #2 (pixels the same as above)
[ RUN ] LayerTreeHostTilesTestPartialInvalidation.PartialRaster_SingleThread_GpuRaster
[24226:24296:0509/105106.024943:2408715788256:ERROR:gles2_cmd_decoder.cc(7783)] [.GpuRasterization]GL ERROR :GL_INVALID_OPERATION : glFramebufferTexture2D: unknown texture_ref
[24226:24611:0509/105106.025226:2408715788543:ERROR:gpu_raster_buffer_provider.cc(54)] Failed to allocate raster surface
[24226:24296:0509/105106.104554:2408715867864:ERROR:gles2_cmd_decoder.cc(7790)] [.GpuRasterization]GL ERROR :GL_INVALID_OPERATION : glFramebufferTexture2D: Attachment textarget doesn't match texture target
[24226:24611:0509/105106.104744:2408715868046:ERROR:gpu_raster_buffer_provider.cc(54)] Failed to allocate raster surface
[24226:24226:0509/105106.437862:2408716201180:ERROR:pixel_comparator.cc(47)] Number of pixel with an error: 30000
[24226:24226:0509/105106.438000:2408716201306:ERROR:pixel_comparator.cc(48)] Error Bounding Box : 0,0 200x200
[24226:24226:0509/105106.460812:2408716224128:ERROR:pixel_test_utils.cc(79)] Pixels do not match!
../../cc/test/layer_tree_pixel_test.cc:117: Failure
Value of: MatchesPNGFile(*result_bitmap_, ref_file_path, *pixel_comparator_)
Actual: false
Expected: true
[ FAILED ] LayerTreeHostTilesTestPartialInvalidation.PartialRaster_SingleThread_GpuRaster (510 ms)
This doesn't appear to be flaking on the bot, and only happened pretty rarely. I had to run like 10 versions of that script while compiling and only saw it happen twice. So, not disabling the test, but it still seems possible that we've got some sort of race in terms of resource creation (maybe flushing?) for gpu partial raster.
|
Deleted:
expected.png
629 bytes
|
|
Deleted:
actual.png
624 bytes
|
Comment 1 by danakj@chromium.org
, Oct 13 2017Status: Duplicate (was: Available)