See http://www.hixie.ch/tests/adhoc/css/overflow/002.html It seems we are now treating overflow:scroll boxes as if they formed a stacking context, which is wrong per 2.1. We should fix this for 1.5.

Just FYI, this regressed between 2005-05-30 06:00 and 2005-05-31 06:00.

I'm almost 100% sure it's because I started creating views for scrollframes in bug 240276, which is now triggering bug 191830. I don't think this can be fixed for 1.5. Fixing 191830 is much too much for branch (although I plan to work on it soon for trunk). Removing the scrollframe view would require some tricky work that would be timeconsuming and risky.

*sigh* make that 2005-03-30 and 2005-03-31 (instead of 2005-05-30 and 2005-05-31).

(Frank: Thanks!) roc: Fair enough. Pity. We really should have caught this earlier; what's our story on regression testing for catching this kind of thing?

Frank says it's actually bug 288117. That's where I introduced the view for scrollframes, so my analysis still holds.

We don't run systematic periodic layout regression tests. They probably would have caught this bug if we had, since it's in your test suite, but my experience has been (from when I was running automated tests) that the user testing we get from nightly builds generally finds more regressions than automatic testing with the test suites we have.

The two aren't mutually exclusive. :-) The more coverage the better.

Yeah, I need to rescuscitate my regression testing scripts again ... and fix the bugs that made painting nondeterministic on some of your testcases.

I think this is WFM in my debug build that has the "frame display lists" patch in it (bug 317375).

Testcase doesn't work with Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.9a1) Gecko/20060124 Firefox/1.6a1 ID:2006012405 (both boxes red, no white text in top box) Testcase does work with Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.9a1) Gecko/20060126 Firefox/1.6a1 ID:2006012613 (both boxes greem, white 'A' in top box) --> RESOLVED FIXED (by bug 317375)