Bug 208817 – [ iOS and Mac Debug ] fast/canvas/webgl/simulated-vertexAttrib0-invalid-indicies.html is flaky timing out and crashing.

• Status: RESOLVED
• Resolution: FIXED
• Component: WebGL

[Jason Lawrence]

fast/canvas/webgl/simulated-vertexAttrib0-invalid-indicies.html

Description:
This test is flaky timing out and crashing on iOS and Mac. It appears to have a regression point of r258025. I was not able to find a valid crash log or diff.

History:
https://results.webkit.org/?suite=layout-tests&test=fast%2Fcanvas%2Fwebgl%2Fsimulated-vertexAttrib0-invalid-indicies.html

Diff/Crash log:
No crash log found for WebKitTestRunner:96552.

stdout:
Content-Type: text/plain
CONSOLE MESSAGE: line 49: WebGL: INVALID_OPERATION: drawElements: attempt to access out of bounds arrays
PASS: MAX_UINT index was unable to be simulated
PASS: MAX_UINT index did not crash
FAIL: Huge index did not fail validation
PASS: Huge index did not crash

#EOF
#EOF

stderr:
#EOF
Unable to shrink memory footprint of process (5839 MB) below the kill thresold (4096 MB). Killed
#CRASHED - com.apple.WebKit.WebContent.Development (pid 97161)
LEAK: 1 WebProcessPool
LEAK: 1 WebPageProxy

[Radar WebKit Bug Importer]

<rdar://problem/60225353>

[Jason Lawrence]

I have marked this test as passing while this issue is investigated.
https://trac.webkit.org/changeset/258145/webkit

[Jason Lawrence]

I can reproduce the issue with r258141 and r258024 using the command below.

run-webkit-tests --iterations 1000 --exit-after-n-failures 3 --debug --force-f fast/canvas/webgl/simulated-vertexAttrib0-invalid-indicies.html

[6/1000] fast/canvas/webgl/simulated-vertexAttrib0-invalid-indicies.html failed unexpectedly (test timed out, test was not run)
[7/1000] fast/canvas/webgl/simulated-vertexAttrib0-invalid-indicies.html failed unexpectedly (test timed out, test was not run)
[8/1000] fast/canvas/webgl/simulated-vertexAttrib0-invalid-indicies.html failed unexpectedly (test timed out, test was not run)
Exiting early after 3 failures. 3 tests run.


0 tests ran as expected, 3 didn't (997 didn't run):

[Alexey Proskuryakov]

Please revert r258025 (and unskip the test).

[Kenneth Russell]

@ap could we please hold off for a couple more hours? It looks like there is some old code in WebKit for vertex attribute 0 simulation that is causing this flaky crash, and that it should be #ifdef'd out with USE_ANGLE.

[Dean Jackson]

I've temporarily reverted 258025. I have a WIP fix for not-simulating vertex attribute 0. I'll turn them both on in the same patch.

[Dean Jackson]

"Fixed" in r258226

[Kenneth Russell]

Linking to the related bug for tracking purposes.