CSS 3d bugs (windows Chrome/Opera)

[Juz]

Hi, I would be grateful if anyone could spare a few moments to confirm any rendering artifacts in the test pages below. The bug has been seen on windows desktop browsers based upon Chromium 30 so that's Chrome 30 and Opera 17

 

The first page just recreates the bug which appears either horizontally or vertically when elements are rotated into the xz or yz planes and then rotated orthogonally round y and x - so they are effectively 'edge on' and vanish

http://juzzer.f2s.com/csstest2.html

 

 

The second test is based on code which was first posted in another thread testing CSS on Firefox (http://www.html5gamedevs.com/topic/336-z-index-is-overlapping-css-3d-transform-position-in-firefox/) where at the time it was noted that artifacts were seen in Canary, which now seem to have progressed to the stable release.

http://juzzer.f2s.com/csstest.html

 

It seems that the bug appears in the initial layout stage and in the first example, restyling is not enough to cause re-layout.

In the second example clicking '-/+' removes/adds cubes. This does cause a re-layout and the glitches disappear but then there is a second bug where anti-aliasing seems to be cancelled.

 

There is a bug logged for the first problem with the chromium project - https://code.google.com/p/chromium/issues/detail?id=306116&q=css%20artifacts&sort=-modified%20-id&colspec=ID%20Pri%20M%20Iteration%20ReleaseBlock%20Cr%20Status%20Owner%20Summary%20OS%20Modified

 

Thanks in advance for any feedback

 

 

[Juz]

So I am the only person to witness this? That's great!

 

After digging through the source code and a few long debugging sessions (chrome/chromium is big and highly multiprocess/threaded!) the problem lies with some recently introduced antialiasing code. As it turns out, once you've learnt a bit of how chrome works you find that the screen rendering gl shaders are in plain text within the dll and easy to modify with a hex editor.

 

This level of coding is something you miss with html5.