Regression from 641838, we add a relation to the set of relations in RefRelationSet() even if there is 0 targets. This makes orca spew lots of back traces because atleast one target was expected, and otherwise makes using firefox with orca very bad.
Created attachment 554881 [details] [diff] [review] patch
Comment on attachment 554881 [details] [diff] [review] patch r=me by inspection. Please test it out.
(In reply to David Bolter [:davidb] from comment #2) > Comment on attachment 554881 [details] [diff] [review] > patch > > r=me by inspection. Please test it out. yeah, alreadyu did, using it now to :)
We don't have any machinary for testing the platform specific accessibility api code at the moment. Fortunately that code is generally a very thin wrapper.
Not having the test framework isn't a reason to set in-testsuite-. Just keep the question mark.
Comment on attachment 554881 [details] [diff] [review] patch This regression hsould be in aurora two since 641838 landed before we branched but this hadn't yet. This patch is very low risk and and only effects linux with a accessibility app running. We've also found no regressions on trunk from it, so I doubt we'd find any before its released. (In any cse I'm pretty sure this won't introduce any regressions.) In theory we could fix the regrssion by backing out 641838. However that would be a much bigger change, and would would probably involve backing out 678189 and 672515 as well. So I think we really want to take this on aurora. p.s. I really should have asked earlier, but I've been dealing with school
Comment on attachment 554881 [details] [diff] [review] patch Reinstating my r+ which mysteriously evaporated.
Landed on tbsaunde's behalf: http://hg.mozilla.org/releases/mozilla-aurora/rev/2ef302863818