Steps to reproduce (copied from bug 80352 comment 45. I'll add a better testcase later): Go to a logical Bidi page, e.g. http://unicode.org/iuc/iuc10/x-ar.html Open a new tab and use about:config to change bidi.texttype to 3 Return to the previous tab. Expected results: the numbers within the Arabic text should be reversed: 21-01 and 7991 instead of 10-12 and 1997. Actual results: the numbers remain the same unless the page is reloaded. Go to a visual Bidi page, e.g. http://www.cs.bgu.ac.il/~elad/noa_v.html Open a new tab and use about:config to change bidi.texttype to 2 Return to the previous tab. Expected results: all the Hebrew text should be reversed: העונ instead of נועה Actual results: the text remains the same unless the page is reloaded.
Since this blocks bug 441782, this should be targeted for 1.9.0.x if the other one is going to make it on a dot release.
Are these two changes equivalent to the nsPresContext::ClearStyleDataAndReflow() function that was orirignally used when you wrote your patch to bug 80352 (attachment 202512 [details] [diff] [review])? <http://bonsai.mozilla.org/cvsview2.cgi?diff_mode=context&whitespace_mode=show&root=/cvsroot&subdir=mozilla/layout/base&command=DIFF&root=/cvsroot&file=nsPresContext.cpp&rev1=3.341&rev2=3.342#6> <http://bonsai.mozilla.org/cvsview2.cgi?diff_mode=context&whitespace_mode=show&root=/cvsroot&subdir=mozilla/layout/base&command=DIFF&root=/cvsroot&file=nsPresContext.cpp&rev1=3.343&rev2=3.344#10>
The fix for this bug is included in the patch for bug 441782.
FWIW, this regressed between 2007-07-25-04 and 2007-07-26-04, presumably by the changes to ClearStyleDataAndReflow in bug 386640.
Hmm, I'm not sure I understand which part of the patch on bug 386640 has caused this regression. I'm CCing Boris and David to see if they have any ideas.
I think it's basically the same regression as bug 394057.
The mPrefChangePendingNeedsReflow = PR_TRUE in the patch in bug 441782 agrees with comment 6; the change that caused the regression was the assumption that things that did StyleChangedReflow were doing something actually changed style data in a way that would trigger reflow. I fixed that assumption in bug 394057, but it seems I missed some cases (these bugs).
OK then. Marking this as a regression of bug 386640. We can mark this as fixed once bug 441782 lands. In the mean time I'm going to write a unit test for this.
Doesn't really meet the "wanted" criteria (security, stability, regression from maintenance release), but we'll look at a reviewed and baked patch (in the other bug).
Fixed on mozilla-central by bug 441782. roc: how can I go about testing this in an automated test?
Does this need a separate testcase? IIRC if this regressed layout/base/tests/test_bug441782.html would fail.
Yes, I think you're right.
If this bug is covered in another test, then this is in‑testsuite+ :)