Issue metadata
Sign in to add a comment
|
Flaky DCHECK in MessagePumpForUI::OnDelayedLooperCallback on Android NVIDIA GPU.FYI bots |
||||||||||||||||||||
Issue descriptionThis causes flakiness on many GPU tests on this bot. For example: https://ci.chromium.org/p/chromium/builders/luci.chromium.ci/Android%20FYI%20Release%20%28Nexus%209%29/6800 https://ci.chromium.org/p/chromium/builders/luci.chromium.ci/Android%20FYI%20Release%20%28Nexus%209%29/6796 08-07 21:32:54.898 27068 27068 F chromium: [FATAL:message_pump_android.cc(121)] Check failed: ret >= 0 (-1 vs. 0) 08-07 21:32:54.898 27068 27068 F chromium: #00 0x0000007f86ceb157 /data/app/org.chromium.chrome-1/lib/arm64/libchrome.so+0x0000000003aea157 08-07 21:32:54.898 27068 27068 F chromium: #01 0x0000007f86cf7c43 /data/app/org.chromium.chrome-1/lib/arm64/libchrome.so+0x0000000003af6c43 08-07 21:32:54.898 27068 27068 F chromium: #02 0x0000007f86cf7a57 /data/app/org.chromium.chrome-1/lib/arm64/libchrome.so+0x0000000003af6a57 08-07 21:32:54.898 27068 27068 F chromium: #03 0x0000007fa32e5007 /system/lib64/libutils.so+0x000000000001c007 08-07 21:32:54.898 27068 27068 F chromium: #04 0x0000007fa32e526b /system/lib64/libutils.so+0x000000000001c26b 08-07 21:32:54.898 27068 27068 F chromium: #05 0x0000007fa317c74b /system/lib64/libandroid_runtime.so+0x00000000000a674b 08-07 21:32:54.898 27068 27068 F chromium: #06 0x0000000072ebc83f /data/dalvik-cache/arm64/system@framework@boot.oat+0x0000000001ee583f Looks like this was recently added in https://chromium-review.googlesource.com/c/chromium/src/+/751322. mthiesse@, PTAL
,
Aug 7
,
Aug 7
Also seeing this on Android FYI Release (NVIDIA Shield TV): https://ci.chromium.org/p/chromium/builders/luci.chromium.ci/Android%20FYI%20Release%20%28NVIDIA%20Shield%20TV%29/5915
,
Aug 8
Taking a look.
,
Aug 9
This is also happening a lot on less exotic. The VR tests on Pixel 1 XLs with Nougat and Oreo seem to hit it fairly frequently, as well as the AR tests on Pixel 1 XLs with Oreo. The regular instrumentation tests are capable of hitting it, at least on Oreo, although much less frequently.
,
Aug 10
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/7c36083d762cfc04d161e605c29bb95229a0bb7e commit 7c36083d762cfc04d161e605c29bb95229a0bb7e Author: Michael Thiessen <mthiesse@chromium.org> Date: Fri Aug 10 20:24:54 2018 Ignore EAGAIN reading timerfd in message_pump_android. EAGAIN should only happen when no timer has expired, but the function should only be called when a timer expires. And yet, very rarely during tests, we error with EAGAIN. I have no theories as to why this is happening. Possibly a platform bug. Ignoring EAGAIN should be safe, we only read the timer to clear it so that we don't keep getting called back. If it's already cleared then we're fine too. Bug: 872044 Change-Id: I6eb29ab99402d90140bd5b89a26df2aa5e53b1c6 Reviewed-on: https://chromium-review.googlesource.com/1168154 Commit-Queue: Michael Thiessen <mthiesse@chromium.org> Reviewed-by: Gabriel Charette <gab@chromium.org> Cr-Commit-Position: refs/heads/master@{#582315} [modify] https://crrev.com/7c36083d762cfc04d161e605c29bb95229a0bb7e/base/message_loop/message_pump_android.cc
,
Aug 13
|
|||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||
Comment 1 by kbr@chromium.org
, Aug 7