Bug 34258 – [Qt] Implement the display() method of the layout test controller

• Component: WebKit Qt
• Status: RESOLVED FIXED

[Benjamin Poulain]

Currently, LayoutTestController::display() is not implemented in the Qt port. This make it impossible to tests problems related to painting.

[Benjamin Poulain]

Created attachment 47611 [details]
Implement the method

With this patch, the web view is shown.
Calling display multiple time is not a problem, call QWidget::show() again is just ignored.

[WebKit Commit Bot]

Comment on attachment 47611 [details]
Implement the method

Clearing flags on attachment: 47611

Committed r54000: <http://trac.webkit.org/changeset/54000>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Csaba Osztrogonác]

(In reply to comment #3)
> All reviewed patches have been landed.  Closing bug.

Rolled out by http://trac.webkit.org/changeset/54002, because it broke buildbot:
http://build.webkit.org/results/Qt%20Linux%20Release/r54000%20%286648%29/results.html

[Benjamin Poulain]

Created attachment 47691 [details]
Implement the method

Some tests are unstable regarding events. It is not safe to process all events in LayoutTestController::reset().

Since we still need a synchronous paint, I have added painting on a pixmap when display() is called.

[Kenneth Rohde Christiansen]

Comment on attachment 47691 [details]
Implement the method

Sounsd like an OK idea

[WebKit Commit Bot]

Comment on attachment 47691 [details]
Implement the method

Clearing flags on attachment: 47691

Committed r54054: <http://trac.webkit.org/changeset/54054>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Andras Becsi]

(In reply to comment #6)
> (From update of attachment 47691 [details])
> Sounsd like an OK ide

This patch has introduced an annoying effect of flickering pop-up frames and shown webviews when running the test suit.
This is a regression as I see it, because the 12000 Mac tests do not show any of this behaviour. Is there a way to fix this in Qt? Kenneth, what do you think?

[Kenneth Rohde Christiansen]

Benjamin implemented this for forcing paints, and we actually dont need to show the window for doing that, now we paint to a pixmap, so that could be changed.

If the behaviour is different than the other platforms we need to investigate why.

[Benjamin Poulain]

(In reply to comment #10)
> Benjamin implemented this for forcing paints, and we actually dont need to show
> the window for doing that, now we paint to a pixmap, so that could be changed.
> 
> If the behaviour is different than the other platforms we need to investigate
> why.

Showing the window is the only way I found to catch some bugs.

(In reply to comment #9)
> This is a regression as I see it, because the 12000 Mac tests do not show any
> of this behaviour. Is there a way to fix this in Qt? Kenneth, what do you
> think?

On Mac the window is shown off screen. I don't know if that works on all platform(?).

[Benjamin Poulain]

Created attachment 47770 [details]
Move the window offscreen

Here is a patch to move the window offscreen.
It should work on X11, I don't know on the other systems (some window manager might reset the position on screen).

Changes:
+ don't show the window on screen
+ still works with plugins and drag-and-drop
- never get any expose event

Note that on X11, without this patch, you can simply run a second X11 server to run the tests.

[Kenneth Rohde Christiansen]

Comment on attachment 47770 [details]
Move the window offscreen

Fine with me

[Andras Becsi]

(In reply to comment #12)
> Created an attachment (id=47770) [details]
> Move the window offscreen
Thank you for the fast fix.