Issue metadata
Sign in to add a comment
|
smoothness.maps failing on 3 builders |
||||||||||||||||||||||
Issue descriptionsmoothness.maps failing on 3 builders Builders failed on: - Win 10 High-DPI Perf: https://build.chromium.org/p/chromium.perf/builders/Win%2010%20High-DPI%20Perf - Win 7 ATI GPU Perf: https://build.chromium.org/p/chromium.perf/builders/Win%207%20ATI%20GPU%20Perf - Win 7 Nvidia GPU Perf: https://build.chromium.org/p/chromium.perf/builders/Win%207%20Nvidia%20GPU%20Perf https://chromium-swarm.appspot.com/task?id=370628875c3ef710&refresh=10&show_raw=1 is an example task. The screenshot shows a blank white page. Looks like there's no test log; maybe it crashed? Something weird is going on. It's failing in roughly the same way on all 3 windows bots. Same blank white screenshot on all 3. Looks like the test is actually failing.
,
Jun 28 2017
The reason this started failing is that I just turned on these tests in Issue 692663 . These tests should definitely be running correctly on Win7 with NVIDIA GPUs. This WPR archive is run (as a correctness test) on Chromium's CQ. Indeed, if we look, we can see the test passing sometimes, but mostly failing. Here's a passing run: https://luci-milo.appspot.com/buildbot/chromium.perf/Win%207%20Nvidia%20GPU%20Perf/936 https://chromium-swarm.appspot.com/task?id=3708290722a2a410&refresh=10&show_raw=1 and a failing run: https://luci-milo.appspot.com/buildbot/chromium.perf/Win%207%20Nvidia%20GPU%20Perf/935 https://chromium-swarm.appspot.com/task?id=370705655cf65710&refresh=10&show_raw=1 These runs were on the *same* machine, build96-m1: https://chromium-swarm.appspot.com/bot?id=build96-m1&sort_stats=total%3Adesc I'm at a loss here. Again, this WPR runs well on similarly configured machines in Chromium's CQ. The test does run very quickly -- in only a couple of seconds. Could there be a race condition in benchmark measurement for very fast running tests? What should I do? Disable the test again? If so I'm going to close Issue 692663 as WontFix.
,
Jun 28 2017
Is it possible to reproduce this on your local Win machine? IIRC, the win machine on CQ is very different from the bare metal ones.
,
Jul 6 2017
When run locally, the benchmark usually displays a 404 for http://localhost:8000/performance.html . If the tab's reloaded in the launched browser, then the benchmark runs to completion. This is probably the reason the test is failing on the bots. It looks like there's a race condition when setting up the WPR archive for replay on Windows. This seems like a generic problem in Telemetry's harness to me -- Ned, could someone from the Telemetry team help us look into this?
,
Jul 6 2017
xunjieli@ is working on migrating WPR server to Go, I think it's better to probably block this bug on that effort.
,
Jul 7 2017
,
Jul 7 2017
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/0e569358d483e0c2b8eb1a4f4868267589d96d17 commit 0e569358d483e0c2b8eb1a4f4868267589d96d17 Author: Kenneth Russell <kbr@chromium.org> Date: Fri Jul 07 01:59:27 2017 Disable smoothness.maps on Windows again. There's a race condition in the WPR server still preventing it from running reliably. BUG= 692663 , 737657 TBR=nednguyen@google.com NOTRY=true Change-Id: Ic2d0f9d85747958e752705536f85b41823977b71 Reviewed-on: https://chromium-review.googlesource.com/562749 Commit-Queue: Kenneth Russell <kbr@chromium.org> Reviewed-by: Kenneth Russell <kbr@chromium.org> Cr-Commit-Position: refs/heads/master@{#484788} [modify] https://crrev.com/0e569358d483e0c2b8eb1a4f4868267589d96d17/tools/perf/benchmarks/smoothness.py
,
Jul 7 2017
Since this is essentially a duplicate of Issue 692663 , closing this one as WontFix. Re-enabling this benchmark on Windows after the WPR server is rewritten in Go will be tracked under the other bug. |
|||||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||||
Comment 1 by nedngu...@google.com
, Jun 28 2017