Issue metadata
Sign in to add a comment
|
v8.runtimestats.browsing_mobile browser:chrome:newtab failing on webview |
||||||||||||||||||||
Issue descriptionv8.runtimestats.browsing_mobile failing on 2 builders Builders failed on: - Android Nexus5X WebView Perf: https://build.chromium.org/p/chromium.perf/builders/Android%20Nexus5X%20WebView%20Perf - Android Nexus6 WebView Perf: https://build.chromium.org/p/chromium.perf/builders/Android%20Nexus6%20WebView%20Perf
,
Oct 10 2017
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/4d071c63f0d7bc6c3277dca6fee3a165e77a4795 commit 4d071c63f0d7bc6c3277dca6fee3a165e77a4795 Author: Oystein Eftevaag <oysteine@chromium.org> Date: Tue Oct 10 17:54:07 2017 Disable failing newtab story of v8.runtimestats.browsing_mobile benchmark on webview TBR=nednguyen@chromium.org Bug: 773077 Change-Id: Idb235cd1d09e9dd35f15e74d07094bab29528e3d Reviewed-on: https://chromium-review.googlesource.com/707778 Reviewed-by: oysteine <oysteine@chromium.org> Commit-Queue: oysteine <oysteine@chromium.org> Cr-Commit-Position: refs/heads/master@{#507710} [modify] https://crrev.com/4d071c63f0d7bc6c3277dca6fee3a165e77a4795/tools/perf/benchmarks/v8_browsing.py
,
Oct 10 2017
,
Oct 10 2017
,
Oct 11 2017
I think newtab and omnibox should not run on webview perf bots. These are disabled here: https://cs.chromium.org/chromium/src/tools/perf/page_sets/system_health/expectations.py?q=system_health/exp&sq=package:chromium&l=146 I guess this is not working because we have a GetExpectations function in v8.runtimestats.browsing_mobile benchmark. https://cs.chromium.org/chromium/src/tools/perf/benchmarks/v8_browsing.py?q=v8_browsi&sq=package:chromium&l=195 In my understanding we should not have GetExpectations inside the benchmark and it should be controlled from system_health/expectations.py instead. @ned, @rnephew, any suggestions on if we can have GetExpectations in v8.runtimestats.browsing_mobile benchmark?
,
Oct 11 2017
Yeah, the system_health page sets/benchmarks are the odd man out for this because of how they are structured. There is no reason they couldn't have their own expectations kept separate and be removed from the system_health expectations file. The v8 benchmarks using the system health page set already have their own classes, so it doesn't really matter where it is located. Do we want it to match the weirdness of system health since its using its page set, or do we want it to match the other benchmarks.
,
Oct 11 2017
Oh ok. I am fine with either approaches. It will be nice to have a consistent behaviour for mobile and desktop versions. So its better we either let the desktop variants to have their own class or move the mobile to use system_health expectations file. I will wait for feedback from ned.
,
Oct 11 2017
Making a separate GetExpectations for v8.runtimestats.browsing_mobile benchmark SGTM
,
Oct 13 2017
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/2aff6db2e54003949511454f195acf42c3ae357d commit 2aff6db2e54003949511454f195acf42c3ae357d Author: Mythri Alle <mythria@chromium.org> Date: Fri Oct 13 13:42:58 2017 Add GetExpectations class to v8.browsing* benchmarks v8.browsing_desktop, v8.browsing_mobile, v8.runtimestats.browsing_desktop were using the expectations from system_health/expectations.py file. This cl moves them to their own class. Bug: chromium:773077 Change-Id: I049361516407455016ee11558d1d2eb9d8e57abc Reviewed-on: https://chromium-review.googlesource.com/716678 Reviewed-by: rnephew <rnephew@chromium.org> Reviewed-by: Ulan Degenbaev <ulan@chromium.org> Reviewed-by: Ned Nguyen <nednguyen@google.com> Commit-Queue: Mythri Alle <mythria@chromium.org> Cr-Commit-Position: refs/heads/master@{#508690} [modify] https://crrev.com/2aff6db2e54003949511454f195acf42c3ae357d/tools/perf/benchmarks/v8_browsing.py [modify] https://crrev.com/2aff6db2e54003949511454f195acf42c3ae357d/tools/perf/page_sets/system_health/expectations.py
,
Oct 13 2017
|
|||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||
Comment 1 by oysteine@chromium.org
, Oct 9 2017