See url, you need to download the testcase to your computer to get the crash, because of enhanced privileges. This is a follow-up from bug 373586. From comment 23 in that bug: This crashes current branch builds when clicking on the button. You need to run the testcase from chrome. I get a [@ nsTextControlFrame::SetValue] (talkback ID: TB32917810X). It doesn't crash on trunk, I think it's because of this patch. I think this is currently nr. 29 of the hit list of crashes on branch: http://talkback-public.mozilla.org/reports/firefox/FF2004/FF2004-topcrashers.html Smaug, maybe this can be fixed by just a null-check?
I don't think so. I added a weakframe test and it showed that the frame got deleted. Probably something similar what bug 373586 did for ::SetValue is needed.
Created attachment 270554 [details] [diff] [review] possible patch Making the ::SetValue to have similar checks as what are currently on trunk. However with the patch using the testcase, the painting of the page doesn't work correctly after pressing the button. (Painting works ok when loading a new page.) The same painting problem happens on trunk too, so I could just file a new (Non-SSensitive) bug for that. The patch is needed anyway, though.
waiting to give branch approval until this shakes out a bit on trunk and we figure out fixes for the painting regression.
Is there a way to trigger this situation without requiring chrome or UniversalXPConnect privileges? It's clearly using a deleted editor object, but if you need privileged code to get there then it's not an exploit. Is the top-crash in this routine likely the same thing? If so then maybe there is a non-chrome way to exploit this. A lot of the talkbacks for this location die on the same "mEditor->SetFlags(savedFlags);" line I die on (and the rest say line "848" which isn't anywhere near this routine and which I'd chalk up to a glitch if it weren't so common and consistent a location). I guess the talkbacks pointing to the same object dereference makes me worry. Whether we want to risk the regressions for the branch hinges on the answer to the above. How bad is the painting regression? (shouldn't the bug have the regression keyword?).
Comment on attachment 270554 [details] [diff] [review] possible patch approved for 188.8.131.52 and 184.108.40.206, a=dveditz for release-drivers
Verified fixed, using: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:220.127.116.11pre) Gecko/20070709 BonEcho/18.104.22.168pre Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:22.214.171.124pre) Gecko/20070710 Firefox/126.96.36.199pre The testcase doesn't crash anymore after clicking on the button.