Re: Issue 586183 in chromium: "WebglConformance.conformance_textures_image_tex_image_and_sub_image_2d_with_image_rgb_rgb_unsigned_byte" is flaky

[c***@googlecode.com]

Updates:
Cc: ***@chromium.org

Comment #1 on issue 586183 by
***@chromium.org: "WebglConformance.conformance_textures_image_tex_image_and_sub_image_2d_with_image_rgb_rgb_unsigned_byte"
is flaky
https://code.google.com/p/chromium/issues/detail?id=586183

Possibly related to Mac flake
https://build.chromium.org/p/chromium.gpu/builders/Mac%2010.10%20Retina%20Release%20%28AMD%29/builds/2335
--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings

[c***@googlecode.com]

Updates:
Status: Assigned
Owner: ***@chromium.org
Labels: -Sheriff-Chromium

Comment #7 on issue 586183 by
***@chromium.org: "WebglConformance.conformance_textures_image_tex_image_and_sub_image_2d_with_image_rgb_rgb_unsigned_byte"
is flaky
https://code.google.com/p/chromium/issues/detail?id=586183

weiliangc, would you mind disabling this test like you wrote in #2?
jmadill, feel free to take over the bug if you'd like to investigate and
fix the issue (which would be even better of course).
--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings

[c***@googlecode.com]

Updates:
Cc: ***@chromium.org ***@chromium.org
Labels: Cr-Blink-WebGL Cr-Blink-Canvas

Comment #10 on issue 586183 by
***@chromium.org: "WebglConformance.conformance_textures_image_tex_image_and_sub_image_2d_with_image_rgb_rgb_unsigned_byte"
is flaky
https://code.google.com/p/chromium/issues/detail?id=586183

Please don't unilaterally disable the test, but try marking it flaky first.

We need to understand what part of the test is timing out. Enne, as GPU
pixel wrangler would you please help by running the test repeatedly
locally, and trying to understand what part is timing out?
--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings

[c***@googlecode.com]

Comment #12 on issue 586183 by
***@chromium.org: "WebglConformance.conformance_textures_image_tex_image_and_sub_image_2d_with_image_rgb_rgb_unsigned_byte"
is flaky
https://code.google.com/p/chromium/issues/detail?id=586183#c12

The following revision refers to this bug:

https://chromium.googlesource.com/chromium/src.git/+/44b2e12bf83d0fa9688847da58525283e850d6fa

commit 44b2e12bf83d0fa9688847da58525283e850d6fa
Author: enne <***@chromium.org>
Date: Wed Feb 17 02:36:42 2016

Mark flaky webgl conformance test as flaky

tex_image_and_sub_image_2d_with_image_rgb_rgb_unsigned_byte appears to
be flaking on all platforms, so mark it as flaky.

BUG=586183
R=***@chromium.org,***@chromium.org

Review URL: https://codereview.chromium.org/1701973004

Cr-Commit-Position: refs/heads/master@{#375765}

[modify]
http://crrev.com/44b2e12bf83d0fa9688847da58525283e850d6fa/content/test/gpu/gpu_tests/webgl_conformance_expectations.py
--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings

[c***@googlecode.com]

Comment #15 on issue 586183 by
***@chromium.org: "WebglConformance.conformance_textures_image_tex_image_and_sub_image_2d_with_image_rgb_rgb_unsigned_byte"
is flaky
https://code.google.com/p/chromium/issues/detail?id=586183#c15

The following revision refers to this bug:

https://chromium.googlesource.com/chromium/src.git/+/da7e8ca6c5966530701a0396ae196b23d5284407

commit da7e8ca6c5966530701a0396ae196b23d5284407
Author: enne <***@chromium.org>
Date: Wed Feb 17 21:08:52 2016

Mark failing android webgl conformance test

tex_image_and_sub_image_2d_with_image_rgb_rgb_unsigned_byte is failing
on android, vs flaky timeouts on other platforms, so mark it as such.

R=***@chromium.org,***@chromium.org
BUG=586183

Review URL: https://codereview.chromium.org/1704093002

Cr-Commit-Position: refs/heads/master@{#375979}

[modify]
http://crrev.com/da7e8ca6c5966530701a0396ae196b23d5284407/content/test/gpu/gpu_tests/webgl_conformance_expectations.py
--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings

[c***@googlecode.com]

Comment #17 on issue 586183 by
***@chromium.org: "WebglConformance.conformance_textures_image_tex_image_and_sub_image_2d_with_image_rgb_rgb_unsigned_byte"
is flaky
https://code.google.com/p/chromium/issues/detail?id=586183

There's no milestone set for this Pri-1 bug. enne, can you please set the
milestone? NOTE: M-49 stable cut is Mar 1.
--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings

[c***@googlecode.com]

Updates:
Cc: ***@chromium.org ***@opera.com ***@chromium.org
Labels: Cr-Blink-MemoryAllocator-GarbageCollection Cr-Blink-Loader

Comment #18 on issue 586183 by
***@chromium.org: "WebglConformance.conformance_textures_image_tex_image_and_sub_image_2d_with_image_rgb_rgb_unsigned_byte"
is flaky
https://code.google.com/p/chromium/issues/detail?id=586183

haraken/sigbjornf/japhet: I have a feeling there is a bug in the lifetime
of HTMLImageElement instances that are created from JavaScript via the "new
Image(...)" constructor.

I suspect there is a brief window when the Image is instantiated and its
src is set, but before it starts loading, where the instance is not
preserved from garbage collection. At that point, if the Image is no longer
referenced from JS and a GC is performed, the instance will be GC'd, and
its onload handler will never be called.

Can any of you three please help dig up the associated code and help
confirm or deny this hypothesis?
--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings

[c***@googlecode.com]

Comment #19 on issue 586183 by
***@opera.com: "WebglConformance.conformance_textures_image_tex_image_and_sub_image_2d_with_image_rgb_rgb_unsigned_byte"
is flaky
https://code.google.com/p/chromium/issues/detail?id=586183

We do have extended lifetime handling in place for such loading, but
detached, image elements (which the named constructor "Image" turns into),
with the v8 GC marking the wrapper object as alive if
HTMLImageElement::hasPendingActivity() is true. Which it will be if the
associated loader has pending activity. It could be that its logic isn't
correct.

However, if I add ASSERT(hasPendingActivity()); after srcAttr is set for
HTMLImageElement, it doesn't reproduce for this conformance test (and
others.)
--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings

[c***@googlecode.com]

Comment #20 on issue 586183 by
***@google.com: "WebglConformance.conformance_textures_image_tex_image_and_sub_image_2d_with_image_rgb_rgb_unsigned_byte"
is flaky
https://code.google.com/p/chromium/issues/detail?id=586183

(First of all, this would be a bug of a V8 GC, not an Oilpan's GC.)

As Sigbjorn pointed out, this would be an issue of
HTMLImageElement::hasPendingActivity. V8GCController checks if
HTMLImageElement::hasPendingActivity returns true and collects the wrapper
only if it returns false. So we might want to check a condition where
HTMLImageElement::hasPendingActivity returns true.
--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings