Bug 133803 – REGRESSION(169799): [GTK] several layout tests on editing/ fail

• Component: Tools / Tests
• Status: RESOLVED FIXED

[Carlos Alberto Lopez Perez]

The following layout test are failing on GTK+ since r169799 <http://trac.webkit.org/changeset/169799>

  editing/deleting/delete-by-word-001.html [ Failure ]
  editing/deleting/delete-by-word-002.html [ Failure ]
  editing/input/option-page-up-down.html [ Failure ]
  editing/input/scroll-viewport-page-up-down.html [ Failure ]
  editing/undo/undo-deleteWord.html [ Failure ]

The diffs are the following:

  http://build.webkit.org/results/GTK%20Linux%2064-bit%20Release%20%28Tests%29/r169799%20%28745%29/editing/deleting/delete-by-word-001-pretty-diff.html
  http://build.webkit.org/results/GTK%20Linux%2064-bit%20Release%20%28Tests%29/r169799%20%28745%29/editing/deleting/delete-by-word-002-pretty-diff.html
  http://build.webkit.org/results/GTK%20Linux%2064-bit%20Release%20%28Tests%29/r169799%20%28745%29/editing/input/option-page-up-down-pretty-diff.html
  http://build.webkit.org/results/GTK%20Linux%2064-bit%20Release%20%28Tests%29/r169799%20%28745%29/editing/input/scroll-viewport-page-up-down-pretty-diff.html
  http://build.webkit.org/results/GTK%20Linux%2064-bit%20Release%20%28Tests%29/r169799%20%28745%29/editing/undo/undo-deleteWord-pretty-diff.html

[Carlos Alberto Lopez Perez]

I added this tests to the GTK TestExpectations on http://webkit.org/b/133811

[Carlos Garcia Campos]

I'm not sure this is a regression, are you sure our own expectation files are correct? The actual result I see for editing/deleting/delete-by-word-001 looks the same than cross-platform expected results. Note that the test does:

    if (navigator.userAgent.search(/\bMac OS X\b/) != -1)
        deleteWordModifiers = ["altKey"];
    else
        deleteWordModifiers = ["ctrlKey"];

So, before r169799 we were using the alt key instead of the control key, which I think was wrong.

[Carlos Alberto Lopez Perez]

(In reply to comment #2)
> I'm not sure this is a regression, are you sure our own expectation files are correct? The actual result I see for editing/deleting/delete-by-word-001 looks the same than cross-platform expected results. Note that the test does:
> 
>     if (navigator.userAgent.search(/\bMac OS X\b/) != -1)
>         deleteWordModifiers = ["altKey"];
>     else
>         deleteWordModifiers = ["ctrlKey"];
> 
> So, before r169799 we were using the alt key instead of the control key, which I think was wrong.

It may be be that the test expectations for this tests before r169799 were wrong.

Lorenzo commented that r169799 made a bunch of tests to pass, so maybe this ones that started to fail just need to be rebaselined.

Compare the results:

169798 -> layout-test 16 failures 13 new passes 23 flakes 1 missing results http://build.webkit.org/builders/GTK%20Linux%2064-bit%20Release%20%28Tests%29/builds/744
169799 -> layout-test 20 failures 21 new passes 15 flakes 1 missing results http://build.webkit.org/builders/GTK%20Linux%2064-bit%20Release%20%28Tests%29/builds/745

[Sergio Villar Senin]

(In reply to comment #3)
> It may be be that the test expectations for this tests before r169799 were wrong.
> 
> Lorenzo commented that r169799 made a bunch of tests to pass, so maybe this ones that started to fail just need to be rebaselined.

Let's rebaseline them and remove as many platform specific results as possible.

[Diego Pino]

The test(s) filed under this bug are consistenly passing since 275882@main (updated baselines for these tests).

https://results.webkit.org/?platform=GTK&platform=WPE&suite=layout-tests&test=editing%2Fdeleting%2Fdelete-by-word-001.html

After updating test expectations in 279290@main, there are no more tests referring to this bug. Closing bug.

[Radar WebKit Bug Importer]

<rdar://problem/128921894>