Bug 182283 – [JHBuild] Add gstreamer-vaapi in moduleset

• Status: RESOLVED
• Resolution: FIXED
• Component: Platform

[Philippe Normand]

.

[Philippe Normand]

Created attachment 332636 [details]
Patch

[EWS Watchlist]

Comment on attachment 332636 [details]
Patch

Attachment 332636 [details] did not pass win-ews (win):
Output: http://webkit-queues.webkit.org/results/6259247

New failing tests:
http/tests/preload/onerror_event.html

[EWS Watchlist]

Created attachment 332644 [details]
Archive of layout-test-results from ews204 for win-future

The attached test failures were seen while running run-webkit-tests on the win-ews.
Bot: ews204  Port: win-future  Platform: CYGWIN_NT-6.1-2.9.0-0.318-5-3-x86_64-64bit

[Philippe Normand]

Created attachment 332654 [details]
Patch

[Carlos Alberto Lopez Perez]

Comment on attachment 332654 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=332654&action=review

> Tools/Scripts/webkitpy/port/gtk.py:120
> +        # Make sure va-api support gets disabled because it's incompatible with Mesa's softGL driver.
> +        environment['LIBVA_DRIVERS_PATH'] = "/tmp"

Likely a minor issue, but I'm a bit uncomfortable with this.
This means each time I run layout tests, something will try to load some library from /tmp that won't exist, ins't it?
But it also means that (theoretically) its possible for an attacker to put in /tmp some crafted library and trick me into executing it.
Instead of /tmp can you look for another directory that is not world-writable? Perhaps another way of disabling this is setting "LIBVA_DRIVER_NAME=null"

[Philippe Normand]

Created attachment 332659 [details]
Patch

[Carlos Alberto Lopez Perez]

Comment on attachment 332659 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=332659&action=review

> Tools/Scripts/webkitpy/port/gtk.py:120
>      def setup_environ_for_server(self, server_name=None):
>          environment = super(GtkPort, self).setup_environ_for_server(server_name)
> +        # Make sure va-api support gets disabled because it's incompatible with Mesa's softGL driver.
> +        environment['LIBVA_DRIVER_NAME'] = "null"

I think this is the wrong place to set this environment variable.
Because we should only do that when the driver is either xvfb or weston (which are the ones using mesa software render). 
The right place would be this one: https://trac.webkit.org/browser/trunk/Tools/Scripts/webkitpy/port/gtk.py?rev=227427#L132

[Philippe Normand]

Created attachment 332663 [details]
Patch

[Carlos Alberto Lopez Perez]

Comment on attachment 332663 [details]
Patch

Interesting mix of spaces we had there. Thanks!

[Carlos Garcia Campos]

Why are we adding this to the jhbuild if it's going to be disabled when running tests?

[Philippe Normand]

(In reply to Carlos Garcia Campos from comment #10)
> Why are we adding this to the jhbuild if it's going to be disabled when
> running tests?

for MiniBrowser testing, mostly.

[Philippe Normand]

Committed r227901: <https://trac.webkit.org/changeset/227901>

[Radar WebKit Bug Importer]

<rdar://problem/37070671>