[GTK] drag&drop related failures

[Philippe Normand]

I haven't yet figured out this regression. But we need green bots so I'll flag the following tests.

Unexpected flakiness: text diff mismatch

  editing/pasteboard/4947130.html = TEXT PASS
  editing/pasteboard/dataTransfer-setData-getData.html = TEXT PASS
  editing/pasteboard/drag-and-drop-image-contenteditable.html = TEXT PASS
  editing/pasteboard/drag-and-drop-inputimage-contenteditable.html = TEXT PASS
  editing/pasteboard/drag-and-drop-objectimage-contenteditable.html = TEXT PASS
  editing/pasteboard/drag-drop-input-textarea.html = TEXT PASS
  editing/pasteboard/drag-drop-modifies-page.html = TEXT PASS
  editing/pasteboard/drag-drop-url-text.html = TEXT PASS
  editing/pasteboard/drag-image-in-about-blank-frame.html = TEXT PASS
  editing/pasteboard/drag-selected-image-to-contenteditable.html = TEXT PASS
  editing/pasteboard/drop-inputtext-acquires-style.html = TEXT PASS
  editing/pasteboard/drop-link.html = TEXT PASS
  editing/pasteboard/drop-text-events.html = TEXT PASS
  editing/pasteboard/drop-text-without-selection.html = TEXT PASS
  editing/pasteboard/file-drag-to-editable.html = TEXT PASS
  editing/pasteboard/file-input-files-access.html = TEXT PASS
  editing/pasteboard/get-data-text-plain-drop.html = TEXT PASS
  editing/pasteboard/subframe-dragndrop-1.html = TEXT PASS

Regressions: Unexpected text diff mismatch :

  fast/forms/drag-into-textarea.html = TEXT
  fast/forms/drag-out-of-textarea.html = TEXT
  fast/forms/file/get-file-upload.html = TEXT
  fast/forms/file/input-file-re-render.html = TEXT
  fast/forms/file/recover-file-input-in-unposted-form.html = TEXT

[Zan Dobersek]

After bisecting the offending commit seems to be r114788.
http://trac.webkit.org/changeset/114788

[Philippe Normand]

(In reply to comment #1)
> After bisecting the offending commit seems to be r114788.
> http://trac.webkit.org/changeset/114788

Really? That seems a bit odd :( I'll have a look at this issue.

[Philippe Normand]

Indeed, spot-on, Zan! I'll try to figure this one out.

[Zan Dobersek]

(In reply to comment #3)
> Indeed, spot-on, Zan! I'll try to figure this one out.

Great, thanks!

[Philippe Normand]

It seems that using this new option --skip-pixel-test-if-no-baseline somehow fixes the issue... Can you confirm Zan?

Not sure about using this on the bots though.

[Philippe Normand]

CCing Dirk and Nandor on this issue. Do we need to use this new --skip-pixel-test-if-no-baseline  option? Any feedback would be appreciated!

[Philippe Normand]

Please also note that GTK bots don't run the pixel tests.

[Dirk Pranke]

I don't know why --skip-pixel-test-if-no-baseline would be causing TEXT regressions; that doesn't make a lot of sense.

If you don't run pixel tests, you shouldn't need to use this option at all.

[Philippe Normand]

(In reply to comment #8)
> I don't know why --skip-pixel-test-if-no-baseline would be causing TEXT regressions; that doesn't make a lot of sense.
> 
> If you don't run pixel tests, you shouldn't need to use this option at all.

Yeah that's what I thought. I'll double-check this tomorrow, maybe a little bug sneaked in the patch introducing that new option, for ports not running pixel tests.

[Zan Dobersek]

Created attachment 139800 [details]
Patch

[Philippe Normand]

Comment on attachment 139800 [details]
Patch

Alright the change and its explanation make sense to me. Thanks!

[Zan Dobersek]

Comment on attachment 139800 [details]
Patch

Clearing flags on attachment: 139800

Committed r115825: <http://trac.webkit.org/changeset/115825>

[Zan Dobersek]

All reviewed patches have been landed.  Closing bug.

[Martin Robinson]

Comment on attachment 139800 [details]
Patch

Awesome fix!