For bug 519486, we're considering not setting the lwthemetextcolor attribute if we can't decide whether a color should be considered bright or dark. But apparently GetDocumentLWTheme from <http://mxr.mozilla.org/mozilla-central/source/content/xul/document/src/nsXULDocument.cpp#4763> makes :-moz-lwtheme depend on the lwthemetextcolor attribute. It should only depend on the lwtheme attribute, as described in bug 513461 comment 0.

taking. Does this have to block 1.9.2 or is this for future updates?

Whoops - sorry, I missed comment 1. You get to flay me if you wish :)

No prob (although I'll take you up on the flaying) ;)

This *might* block 1.9.2 - depends on if we end up needing it for bug 519486 (which I'm working on right now). Either way, it'd be great to have on 1.9.2 and is a low-risk fix.

Comment on attachment 412516 [details] [diff] [review] Patch v1 r=dbaron (I thought the original idea was that the lwthemetextcolor attribute was required, but I guess that's not true anymore.)

(In reply to comment #6) > (I thought the original idea was that the lwthemetextcolor attribute was > required, No, that must have been a misunderstanding.

This may have caused a 0.31% Tp4 regression on Linux: <http://graphs.mozilla.org/graph.html#tests=[{"machine":188,"test":38,"branch":1},{"machine":189,"test":38,"branch":1},{"machine":190,"test":38,"branch":1},{"machine":191,"test":38,"branch":1},{"machine":192,"test":38,"branch":1},{"machine":193,"test":38,"branch":1},{"machine":194,"test":38,"branch":1},{"machine":195,"test":38,"branch":1},{"machine":196,"test":38,"branch":1},{"machine":197,"test":38,"branch":1},{"machine":198,"test":38,"branch":1},{"machine":199,"test":38,"branch":1},{"machine":200,"test":38,"branch":1},{"machine":201,"test":38,"branch":1},{"machine":202,"test":38,"branch":1},{"machine":203,"test":38,"branch":1},{"machine":204,"test":38,"branch":1},{"machine":205,"test":38,"branch":1}]&amp;sel=1258277040,1258449840> The graph looks noisy, so this might not be a real regression after all.

(In reply to comment #9) > This may have caused a 0.31% Tp4 regression on Linux: > > <http://graphs.mozilla.org/graph.html#tests=[{"machine":188,"test":38,"branch":1},{"machine":189,"test":38,"branch":1},{"machine":190,"test":38,"branch":1},{"machine":191,"test":38,"branch":1},{"machine":192,"test":38,"branch":1},{"machine":193,"test":38,"branch":1},{"machine":194,"test":38,"branch":1},{"machine":195,"test":38,"branch":1},{"machine":196,"test":38,"branch":1},{"machine":197,"test":38,"branch":1},{"machine":198,"test":38,"branch":1},{"machine":199,"test":38,"branch":1},{"machine":200,"test":38,"branch":1},{"machine":201,"test":38,"branch":1},{"machine":202,"test":38,"branch":1},{"machine":203,"test":38,"branch":1},{"machine":204,"test":38,"branch":1},{"machine":205,"test":38,"branch":1}]&amp;sel=1258277040,1258449840> > > The graph looks noisy, so this might not be a real regression after all. Yea, I'd put it down to noise (or something else) - the only real code change here is a single enum assignment.

Comment on attachment 412516 [details] [diff] [review] Patch v1 Approved fragility fix.