Bug 60236 – full-screen-stacking-context-expected.txt should be under platform/mac

• Status: RESOLVED
• Resolution: FIXED
• Component: Tools / Tests

[Hajime Morrita]

A test result commited at r85811 contains the layout dump, which is platform dependent.
We should move it to platform directly.

[Hajime Morrita]

Created attachment 92356 [details]
Patch

[Kent Tamura]

Comment on attachment 92356 [details]
Patch

The patch has only ChangeLog.

[Hajime Morrita]

Hi Kent-san, thank you for taking a look!

(In reply to comment #2)
> (From update of attachment 92356 [details])
> The patch has only ChangeLog.
Hmm. It seems webkit-patch post doesn't support simple file move...
I didn't know that.

[Jer Noble]

(In reply to comment #0)
> A test result commited at r85811 contains the layout dump, which is platform dependent.
> We should move it to platform directly.

Why is the layout dump platform dependant?  The pixel test, perhaps, but the layout itself should be platform agnostic.

[Jer Noble]

Note that other tests in the LayoutTests/fullscreen/ directory have pixel test results in LayoutTests/platform/mac/fullscreen/ and layout dumps within LayoutTests/fullscreen/, notably the full-screen-zIndex.html test.

[Tony Chang]

(In reply to comment #4)
> Why is the layout dump platform dependant?  The pixel test, perhaps, but the layout itself should be platform agnostic.

Any test that has text output will likely differ by platforms since the font metrics are different.

For example, full-screen-zIndex.html has separate -expected.txt results for chromium-win, chromium-linux and gtk.  The test is skipped on win and qt.

[Jer Noble]

(In reply to comment #6)
> (In reply to comment #4)
> > Why is the layout dump platform dependant?  The pixel test, perhaps, but the layout itself should be platform agnostic.
> 
> Any test that has text output will likely differ by platforms since the font metrics are different.

Ah, I see, so the bounds of each text line dumped will be different.  Should these tests be re-written to remove the text portions?

[Tony Chang]

(In reply to comment #7)
> (In reply to comment #6)
> > (In reply to comment #4)
> > > Why is the layout dump platform dependant?  The pixel test, perhaps, but the layout itself should be platform agnostic.
> > 
> > Any test that has text output will likely differ by platforms since the font metrics are different.
> 
> Ah, I see, so the bounds of each text line dumped will be different.  Should these tests be re-written to remove the text portions?

Another option is to use layoutTestController.dumpAsText(true).  This tells DRT to use the dumpAsText output + include a pixel result.

[Alexey Proskuryakov]

> Should these tests be re-written to remove the text portions?

If that can be done without making the tests harder to run manually, and the tests can't be text-only, then that's definitely the best solution.

My dream is to have a font that's rendered by all platforms identically :)

[Jer Noble]

(In reply to comment #8)
> Another option is to use layoutTestController.dumpAsText(true).  This tells DRT to use the dumpAsText output + include a pixel result.

This seems like the optimal scenario.  I'll look into doing this.

[Jer Noble]

Created attachment 92435 [details]
Patch

[Tony Chang]

Comment on attachment 92435 [details]
Patch

Can you delete the other platform/*/fullscreen/full-screen-zIndex-expected.txt files in this patch?  Otherwise, this looks good.

[Jer Noble]

Created attachment 92458 [details]
Patch

[Tony Chang]

Comment on attachment 92458 [details]
Patch

Thanks!

[Jer Noble]

Comment on attachment 92458 [details]
Patch

Adding cq+.

[WebKit Commit Bot]

The commit-queue encountered the following flaky tests while processing attachment 92458 [details]:

java/lc3/JavaObject/JavaObjectToLong-005.html bug 60309 (author: [e-mail])
The commit-queue is continuing to process your patch.

[WebKit Commit Bot]

Comment on attachment 92458 [details]
Patch

Clearing flags on attachment: 92458

Committed r85897: <http://trac.webkit.org/changeset/85897>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[WebKit Review Bot]

http://trac.webkit.org/changeset/85897 might have broken GTK Linux 32-bit Release
The following tests are not passing:
fullscreen/full-screen-stacking-context.html