Bug 160057 - Benchmark test content-animation timeouts always
Summary: Benchmark test content-animation timeouts always
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: Tools / Tests (show other bugs)
Version: WebKit Nightly Build
Hardware: Unspecified Unspecified
: P2 Normal
Assignee: Carlos Alberto Lopez Perez
URL:
Keywords:
Depends on:
Blocks: 156827
  Show dependency treegraph
 
Reported: 2016-07-21 15:51 PDT by Carlos Alberto Lopez Perez
Modified: 2016-08-03 11:07 PDT (History)
10 users (show)

See Also:


Attachments
Patch (2.53 KB, patch)
2016-08-03 10:55 PDT, Carlos Alberto Lopez Perez
no flags Details | Formatted Diff | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Carlos Alberto Lopez Perez 2016-07-21 15:51:16 PDT
The benchmark test content-animation added in https://trac.webkit.org/changeset/199805 timeouts.

The GTK+ performance bots is running all the benchmark tests plans, and this is one is constantly timing out. Check: https://build.webkit.org/builders/GTK%20Linux%2064-bit%20Release%20%28Perf%29 It uses the "--allplans" command line switch that runs all *.plan one after other.

I tried running it on the GTK+ platform:

$ Tools/Scripts/run-benchmark --platform gtk --browser minibrowser --plan content-animation

and it timeouts

I tried running it also on Mac platform:

$ Tools/Scripts/run-benchmark --platform osx --browser safari --plan content-animation

and it also timeouts.

... looking at the changelog of r199805 it says:

"""
When run as a benchmark, test completion involves a couple of bounces between the test and MobileSafari.
Test completion is initiated by a setTimeout() in the test, at which point location.hash is set to "#done".
MobileSafari detects that, and asynchronously dumps performance data. Once this is complete, MobileSafari
changes the URL has to "#submit", which triggers the XHR to the test relay that indicates the test is complete.
"""

So, if I understand things correctly, this tests depends on some black-magic that only happens on MobileSafari to work?

This is very unfortunate.

I think we should either:

Option 1) Fix this test so it can run with any browser.
Option 2) Add a way to skip test plans by platform/browser or make some tests to only run on some very specific platform/browser.

Opinions?
Comment 1 Simon Fraser (smfr) 2016-07-21 16:01:54 PDT
I thought I skipped them but can't find that commit now.
Comment 2 Simon Fraser (smfr) 2016-07-21 16:04:13 PDT
Ah, I did in https://trac.webkit.org/changeset/199906
Comment 3 Carlos Alberto Lopez Perez 2016-07-21 16:10:55 PDT
(In reply to comment #2)
> Ah, I did in https://trac.webkit.org/changeset/199906

That is for performance tests (that are run with the Script Tools/Scripts/run-perf-tests

Here I talk about the benchmark test that is ran with the script Tools/Scripts/run-benchmark

The GTK perf test bot runs both kind of performance tests. It has 2 steps. First runs perf-tests and after that it runs benchmark-tests
Comment 4 Carlos Alberto Lopez Perez 2016-08-03 10:55:29 PDT
Created attachment 285248 [details]
Patch
Comment 5 Carlos Alberto Lopez Perez 2016-08-03 11:07:14 PDT
Comment on attachment 285248 [details]
Patch

Clearing flags on attachment: 285248

Committed r204087: <http://trac.webkit.org/changeset/204087>
Comment 6 Carlos Alberto Lopez Perez 2016-08-03 11:07:23 PDT
All reviewed patches have been landed.  Closing bug.