Bug 33936 – [GTK] Fails /LayoutTests/fast/css/button-height.html

• Component: Tools / Tests
• Status: RESOLVED FIXED

[Daniel Bates]

The GTK port fails test /LayoutTests/fast/css/button-height.html, which was included in the patch for bug #29564.

The GTK bot results are at <http://build.webkit.org/results/GTK%20Linux%20Release/r53591%20%287900%29/fast/css/>.

From the results:

FAIL document.getElementById('button1').offsetHeight should be 25. Was 28.

The GTK port renders <input type="button"> and <button> elements (without any user-specified style) with different heights. The former is 25 pixels tall. The latter is 28 pixels tall.

Moreover, the GTK port does not honor the user-specified height for <input type="button" style="height: 40px" .../> for the fifth button in the test. Hence, the failure message:

FAIL document.getElementById('button5').offsetHeight is incorrect for this platform. Should be the same height as button 1 for the Mac ports and 40 otherwise.

[Daniel Bates]

Rolling out change set 53591 <http://trac.webkit.org/changeset/53591> since it caused a regression of the test /fast/replaced/table-percent-height.html on the Qt bot.

There appears to be an issue on GTK with regards to the rendering of button heights, but since I am rolling out r53591 (which includes the test button-height.html) it invalidates the reason this bug was filed.

We should probably look into the GTK button rendering code/file a separate bug.

[Daniel Bates]

Re-committed patch that includes test button-height.html in change set 53897 <http://trac.webkit.org/changeset/53897>.

[Martin Robinson]

Created attachment 70292 [details]
Patch for this issue

[Daniel Bates]

Comment on attachment 70292 [details]
Patch for this issue

This looks sane to me.

I assume at some place the height of the button is computed based on the height of the button text.

[Martin Robinson]

Committed r69436: <http://trac.webkit.org/changeset/69436>