[Webkit-unassigned] [Bug 104662] [EFL][GTK][Mac][Chromium] Share accessibility LayoutTest

bugzilla-daemon at webkit.org bugzilla-daemon at webkit.org
Fri Feb 1 03:11:37 PST 2013


--- Comment #19 from Mario Sanchez Prada <mario at webkit.org>  2013-02-01 03:13:38 PST ---
(In reply to comment #18)
> (In reply to comment #17)
> > is accessibility the only thing that atk platforms share? it seems like
> > there could be an atk folder in LayoutTests/platform that will hold these
> > types. then ATK type platform would run the tests in their own platform and
> > in Atk
> Yes, ATK is just accessibility thing. It is possible to create
> LayoutTests/platform/atk dir and put test there, but atk is not a platform, 
> and some of us may not like it.

I agree (see comment #15) that atk is not really a platform, and that certainly makes it quite weird if we add a platform/atk directory (we can't do run-webkit-tests --atk after all, neither it would make sense IMHO)

In the other hand, however, atk is certainly not a WebKit feature either but a specific set of interfaces that are only used in some ports, and so I can't avoid feeling like and accessibility/atk directory (even if you make sure it's skipped in every platform but GTK anr EFL) is also weird.

Still I see the obvious gain of reusing the tests for the two platforms using the same a11y code, so it's clear it would be good if we found a solution here...

I personally think that this is a very peculiar case because accessibility itself is different in some ways to some other stuff in WebKit like rendering or multimedia for instance, regardless of the specific differences between platforms in those cases (specific mm backends or graphics libraries). At least in those cases the basic structure of the tests are normally the same (a DOM tree is the same DOM tree regardless of the platform).

However, in accessibility that's not always true: different ports offer different views of the accessibility *hierarchy* from WebCore, by means of the platform-specific a11y hierarchy (e.g. more or less "flattened") represented by all those wrapper objects. And so far this was not a problem because only GTK used that different structure, so it was easy to just put the tests expecting a different a11y hierarchy in the platform/gtk directory. But now the EFL port is also interested in that hierarchy, which is no longer a one-platform-specific thing, yet not a global thing to be considered either.

But still, at least in my opinion, is something leaning more towards the platform-specific worlds under LayoutTests/platform/ than towards the more global world that is just under LayoutTests/, so that's why I proposed that change in the first place: to create a platform/atk directory (yes, even if it's not a "true platform") and modify _search_paths() both in layout_tests/port/gtk.py and layout_tests/port/efl.py so the gtk and efl ports will run those tests, in a similar way to how other directories as gtk-wk1 and gtk-wk2 are loaded for GTK (even if gtk-wk1 and gtk-wk1 are not really "platforms"), or efl-wk1 and efl-wk2 for EFL.

Still, I know the simil is not 100% accurate either and beware I'm not really happy either about the platform/atk thing, but I think that's probably a better solution than just creating a LayoutTests/accessibility/atk directory just for the sake of sharing that stuff between these two platforms.

Anyway, this is not a strong opinion anyway. I'm just trying to be constructive and expose the reasoning behind my suggestion so you better understand why I'm proposing that. But I'm of course open to the other option you proposed (or any other) as long as we all are reasonably happy with the result.

Hope this helps.

Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

More information about the webkit-unassigned mailing list