GNOME Bugzilla – Bug 565886
[verified] Extraneous caret-moved events emitted by FF 3.1 cause unnecessary braille updating
Last modified: 2009-01-13 19:48:49 UTC
At some point in the development of Firefox 3.1, caret-moved events began to be emitted for certain focusable items immediately after they were given focus (e.g. through tabbing or structural navigation). As a result of this extra event, we update braille a second time (once for focus, once for caret-moved). The most significant impact (in my mind) of our handling these caret-moved events by sending them off to the default script is that the regression test output for 3.0 will often be different from 3.1 making it challenging to have a single set of tests which we can use. Beyond this, while I haven't yet done performance testing, updating braille isn't the most performant thing we do. :-) If we update once rather than twice under these circumstances, it certainly couldn't hurt.... :-)
Created attachment 125448 [details] [review] revision 1 Pylinted and regression tested in OpenSolaris (FF 3.0.5 and FF 3.1beta2). This doesn't change things for FF 3.0.5, and it makes the regression tests for 3.1beta2 have fewer unexpected failures due to double-brailling). I also tested this manually in Yelp. I have not yet tested this in Thunderbird (or whatever it's calling itself these days). Please test, as I am working on the regression tests and would like to check in this patch if it's sound. Thanks!
works fine for me in intrepid standard ff3 as well as the nightly ff3 build. Sorry dont use yelp nor Thunderbird. Using a voyager 44
Thanks Jon! I tested with the latest Thunderbird trunk contrib build for OpenSolaris last night and this morning. There do not seem to be any side effects. Given that, and the limited nature of the patch, I've gone ahead and checked it into trunk for wider testing before the 2.25.4 release. Moving to pending.
Created attachment 125665 [details] [review] Do the same fix for ARIA push buttons In going through all of the regression tests, I noticed we're double-braille-updating certain ARIA push buttons. These things are emitting caret-moved events after the focus/focused events. There's no good reason for that. :-) We were already ignoring caret-moved events for ARIA page tab, list item, and menu item. This patch just adds push button. Limited to the Gecko script, already pylinted, regression tested across all three versions of Firefox, and committed to trunk.
This seems safe and appropriate for gnome-2-24 as well. Joanie -- if you agree, please check it into gnome-2-24. Thanks!
Both patches committed to the gnome-2-24 branch. Readjusting the milestone.
I've been using orca heavily with braille over the past couple days and haven't seen any problems related to this.
Thanks Mike! Closing as FIXED.