I'm using Arora, which is a browser based on QtWebKit on Mandriva Linux Cooker, and am trying to use it to render:
Now it has a button called "Show Navigation Controls", which has the following CSS:
border: 2pt solid black;
Now Firefox and git://xn--55d.com/kickass.git (a WebKit and Perl/Gtk based browser) displays the button all crisp and beautiful, but it looks horrible on Arora. It has rounded corners, but they are positively ugly.
I'll attach a screenshot shortly.
-- Shlomi Fish
Created attachment 34397 [details]
The button as it looks in Arora.
This is the button as it looks in Arora. One can see that it looks pixelated, and that the horizontal and vertical radii are uneven.
This appears to be a problem with the Qt rendering backend; it renders correctly when using the raster rendering backend.
Reproduced on Linux.
Renders correctly on Windows.
Created attachment 50400 [details]
The button as it looks in Arora on Windows.
This bug might be related to bug #31644.
Works on Snow Leopard with Qt 4.7 trunk (HEAD 03f8f1df0d88f5ffe0b3120cffce614cbeefdb70) and WebKit trunk (r59155), tested on QtLauncher.
Also works on Linux, tested with '-graphicssystem raster'.
*** Bug 31644 has been marked as a duplicate of this bug. ***
Created attachment 83797 [details]
screenshot of still present defect when using border-radius
This bug is still present, with or without "-graphicssystem raster".
There is a visible difference when using using the raster backend, but the imperfections are still apparent and rounded corners are pixelated. Screenshot shows comparison with Chrome, where borders are smooth and not jagged.
This bug should be re-opened.
http://oserv.org/bugs/qtwebkit-border-radius/test1.html — bad (but seems to depend on the version and enviroment)
http://oserv.org/bugs/qtwebkit-border-radius/test1a.html — good
http://oserv.org/bugs/qtwebkit-border-radius/test2.html — bad (even in Qt 5.1)
http://oserv.org/bugs/qtwebkit-border-radius/test2a.html —- good
Can someone reopen this, please?
Reopen since we still have cases without antialiasing.
Created attachment 209067 [details]
StylePainterQStyle would call StylePainter::init() twice, the second time that would overwrite m_previousAntialiasing so that when the StylePainter is closed we would disable antialiasing.
Created attachment 209069 [details]
Comment on attachment 209069 [details]
Clearing flags on attachment: 209069
Committed r154270: <http://trac.webkit.org/changeset/154270>
All reviewed patches have been landed. Closing bug.
Thank you, I will test my original source (from which the above testcases were extracted) when the new version will be released and packaged for my distribution.
If there will be any rendering glitches, I will make another testcase and post it here.