TestTypingInTextField is extremely flaky on ipad-x64-perf bot |
||||||
Issue descriptionTestTypingInTextField fails more than half of the time on perf bot: https://uberchromegw.corp.google.com/i/internal.bling.main/builders/ipad-x64-perf here is an test failure example: https://logs.chromium.org/v/?s=chrome%2Fbb%2Finternal.bling.main%2Fipad-x64-perf%2F1081%2F%2B%2Frecipes%2Fsteps%2Fios_chrome_perftests__iPad_5th_Gen_iOS_11.2.1__on_iOS-11.2.1%2F0%2Fstdout
,
Jan 5 2018
Issue 798960 has been merged into this issue.
,
Jan 5 2018
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/97d6d383e64e3a0ad1e570a42ccb2ada497e37ee commit 97d6d383e64e3a0ad1e570a42ccb2ada497e37ee Author: Yuke Liao <liaoyuke@chromium.org> Date: Fri Jan 05 17:22:26 2018 Disable TestTypingInTextField on perf bot. TestTypingInTextField is extremely flake as it fails more than half of the runs. TBR=marq@chromium.org Bug: 799488 Cq-Include-Trybots: master.tryserver.chromium.mac:ios-simulator-cronet;master.tryserver.chromium.mac:ios-simulator-full-configs Change-Id: I5447ada195f65d8f68c44a19939525236851e986 Reviewed-on: https://chromium-review.googlesource.com/852513 Reviewed-by: Yuke Liao <liaoyuke@chromium.org> Commit-Queue: Yuke Liao <liaoyuke@chromium.org> Cr-Commit-Position: refs/heads/master@{#527307} [modify] https://crrev.com/97d6d383e64e3a0ad1e570a42ccb2ada497e37ee/ios/chrome/browser/ui/omnibox_perftest.mm
,
Jan 5 2018
,
Jan 8 2018
I looked at this and here's what I found: The test is supposed to type text in the omnibox and time how long it takes for the suggestions to appear. This is achieved by: 1. appending text with setText: 2. waiting for the runloop to spin 3. when the test ends, the runloop is span again, for the typing to "complete" It seems like after CL 844079 (https://chromium-review.googlesource.com/c/chromium/src/+/844079), spinning runloop once is not enough for the suggestions to actually start appearing in the popup. Only the clipping is resolved after the first runloop spin. This leads the test to believe that the characters were typed before they actually change in the textfield. As a result, when the test is resetting by calling setText: with an empty string, it doesn't *actually* reset, because the characters are still being typed. I'll have to investigate a solution for this. The good news is that this is a false positive - typing into the omnibox works just fine. Hence this should not be RBD.
,
Jul 16
,
Jul 16
,
Aug 20
Punting this to M-71 as it seems like this will be much easier to fix if we only care about UI Refresh where we don't have clipping. |
||||||
►
Sign in to add a comment |
||||||
Comment 1 by liaoyuke@chromium.org
, Jan 5 2018