Bug 54617 – REGRESSION: transform:matrix()-based rotation sometimes doesn’t reverse correctly

• Status: RESOLVED
• Resolution: WORKSFORME
• Component: CSS

[Oli Studholme]

in WebKit Version 5.0.3 (6533.19.4, r78685) only matrix-based rotations sometimes don’t reverse correctly when triggering stops (e.g. trigger transition several times while transitioning), leading to animation to incorrect endpoint then instant snap to actual endpoint. This problem is not visible in Safari Version 5.0.3 (6533.19.4) or Chrome 9.0.597.102 — regression?
Linked example works as expected in Firefox 4 beta and Opera 11.01
http://oli.jp/bugs/webkit/transform-matrix-rotate.html

[Chris Marrin]

(In reply to comment #0)
> in WebKit Version 5.0.3 (6533.19.4, r78685) only matrix-based rotations sometimes don’t reverse correctly when triggering stops (e.g. trigger transition several times while transitioning), leading to animation to incorrect endpoint then instant snap to actual endpoint. This problem is not visible in Safari Version 5.0.3 (6533.19.4) or Chrome 9.0.597.102 — regression?
> Linked example works as expected in Firefox 4 beta and Opera 11.01
> http://oli.jp/bugs/webkit/transform-matrix-rotate.html

At first you say 5.0.3 shows the problem, but then you say it doesn't. Where are you seeing results that are different?

The problem is that we have to decompose the matrix to extract the rotations and that process can lose sign. I don't see a way around this. But I am concerned that it behaves differently on different browsers. Let me know if that is the case

[Oli Studholme]

Created attachment 82895 [details]
testcase that can be used to demonstrate transitioning matrix-based rotations reversing incorrectly

screenshot to follow

[Oli Studholme]

Created attachment 82898 [details]
After mouseover & mouseout-ing several timesm matrix-based rotations get confused

@Chris Marrin — *Safari 5.0.3* doesn’t demonstrate this behaviour, *WebKit r78685* does — I also wish they weren’t quite so similar (like WebKit not opening Safari’s tab state on launch ;). Hopefully this screenshot demonstrates the difference between this bug and 54616

[Simon Fraser (smfr)]

I think I fixed this via bug 55022 or bug 54793. Please test and reopen if it still occurs.