Bug 230705 – [GTK] Can't run performance tests due to a11y errors

• Component: Tools / Tests
• Status: RESOLVED FIXED

[Sergio Villar Senin]

$ Tools/Scripts/run-perf-tests PerformanceTests/Layout/flexbox-column-wrap.htmlRunning 1 testsRunning 
WebKit/PerformanceTests/Layout/flexbox-column-wrap.html (1 of 1)
error: WebKit/PerformanceTests/Layout/flexbox-column-wrap.html

(WebKitTestRunner:19): dbind-WARNING **: 11:46:41.779: Couldn't connect to accessibility bus: Failed to connect to socket /tmp/dbus-idgbDFkCBq: No such file or directory

(WebKitWebProcess:51): dbind-WARNING **: 11:46:41.934: Couldn't connect to accessibility bus: Failed to connect to socket /tmp/dbus-idgbDFkCBq: No such file or directory

FAILED
Finished: 5.210619 s

[Sergio Villar Senin]

Might be related to the flatpak sandbox?

[Philippe Normand]

I think the API test runner was recently patched to activate the a11y dbus stuff... Maybe something similar is needed here?

[Carlos Garcia Campos]

No, this is a different error, it's indeed related to the sandbox, it happened to me always when apps using the sandbox where running, in my case evolution and devhelp. I debugged it, but didn't manage to find a solutiuon, however, it got fixed automagically after a sysem upgrade, so I assume the problem was in libdbus or some other related package. See this thread https://bugs.webkit.org/show_bug.cgi?id=229032#c3

[Carlos Garcia Campos]

I don't know why that warning affects performance tests though

[Sergio Villar Senin]

(In reply to Carlos Garcia Campos from comment #4)
> I don't know why that warning affects performance tests though

There should be some return value fooling the run-performance-tests tool

[Carlos Garcia Campos]

(In reply to Sergio Villar Senin from comment #5)
> (In reply to Carlos Garcia Campos from comment #4)
> > I don't know why that warning affects performance tests though
> 
> There should be some return value fooling the run-performance-tests tool

or G_DEBUG=fatal_warnings in the test runner?

[Patrick Griffis]

Created attachment 448403 [details]
Patch

[Patrick Griffis]

This patch doesn't fix all a11y issues but I believe this will fix the one mentioned here.

Also tested by lmoura as working but more testing would be appreciated.

[Patrick Griffis]

Created attachment 448406 [details]
Patch

[Patrick Griffis]

This fails on the runner because it uses `gdbus` but I also can't use pygobject's Gio bindings directly as that is also not installed.

I'm not familiar with the process of adding deps to the runner. We should add pygobject if possible as its generally useful.

[Carlos Garcia Campos]

Is this still a problem after the switch to ATSPI?

[Sergio Villar Senin]

(In reply to Carlos Garcia Campos from comment #11)
> Is this still a problem after the switch to ATSPI?

I've tried with r289814 and it's still failing

[Carlos Garcia Campos]

Created attachment 452495 [details]
Disable a11y in WTR

Sergio, could you try this patch? I really don't understand why it happens with perf tests and not layout tests, since both use WTR, but I think we can just disable a11y in WTR, because it's not really needed now with ATSPI.

[Sergio Villar Senin]

(In reply to Carlos Garcia Campos from comment #13)
> Created attachment 452495 [details]
> Disable a11y in WTR
> 
> Sergio, could you try this patch? I really don't understand why it happens
> with perf tests and not layout tests, since both use WTR, but I think we can
> just disable a11y in WTR, because it's not really needed now with ATSPI.

That patch worked for me, I got perf results.

NOTE: I still got 

** (WebKitWebProcess:251): WARNING **: 02:27:09.234: Can't connect to a11y bus: Could not connect: No such file or directory

but the test run fine

[Carlos Garcia Campos]

(In reply to Sergio Villar Senin from comment #14)
> (In reply to Carlos Garcia Campos from comment #13)
> > Created attachment 452495 [details]
> > Disable a11y in WTR
> > 
> > Sergio, could you try this patch? I really don't understand why it happens
> > with perf tests and not layout tests, since both use WTR, but I think we can
> > just disable a11y in WTR, because it's not really needed now with ATSPI.
> 
> That patch worked for me, I got perf results.
> 
> NOTE: I still got 
> 
> ** (WebKitWebProcess:251): WARNING **: 02:27:09.234: Can't connect to a11y
> bus: Could not connect: No such file or directory
> 
> but the test run fine

This comes from the MiniBrowser showing the results, which confirms the problem is not specific to the perf tests runner.

I think it's a good idea in any case to disable a11y in WTR, because it's not actually needed, but there's still a problem that I don't understand.

[Carlos Garcia Campos]

Created attachment 452719 [details]
Disable a11y in WTR

[Carlos Garcia Campos]

Committed r290293 (?): <https://commits.webkit.org/r290293>

[Radar WebKit Bug Importer]

<rdar://problem/89281902>

[Carlos Garcia Campos]

Comment on attachment 448406 [details]
Patch

Clearing flags. It's possible that we still need this patch, if that's the case we will handle it in a new bug.