You need an account on this bank to reproduce... STEPS TO REPRODUCE: 1. enter URL 2. dialog pops up asking for the master password for Software Security Device 3. enter password (personal certificate is required to log in) 4. dialog disappears and Mozilla crashes before any content from the site appears. (Talkback is not triggered) 100% reproducible. If I choose "Cancel" in the password dialog the crash does not appear, and I am redirected to the "error page" of the site as expected. PLATFORMS & BUILDS TESTED: Bug occurs in Mozilla 2005-04-08-05 on Linux. ADDITIONAL INFORMATION: Regression window 2005-04-05-05 -- 2004-04-06-05, checkins in that period: http://bonsai.mozilla.org/cvsquery.cgi?treeid=default&module=all&branch=HEAD&branchtype=match&dir=&file=&filetype=match&who=&whotype=match&sortby=Date&hours=2&date=explicit&mindate=2005-04-05+05%3A00&maxdate=2005-04-06+05%3A00&cvsroot=%2Fcvsroot I will cvs update a debug tree and try to get a stack...

Results from a Linux debug build: Nothing happens I type ENTER in the password dialog, the dialog stays up and nothing is loaded. If I click OK instead then I can log in successfully. I see this assertion after I entered the URL but before the password dialog is presented: ###!!! ASSERTION: wrong thread: 'PR_GetCurrentThread() == gSocketThread', file nsHttpConnection.cpp, line 789 If I use the OK button in the 2005-04-08-05 nightly build the crash does not occur and I can log in, so the bug only seems to occur when typing ENTER.

FWIW, this is the crash stack from a 2005-04-08-05 nightly build.

There's a good chance this is a dup of bug 289379. Does talkback trigger on the nightly crash?

(In reply to comment #3) > There's a good chance this is a dup of bug 289379. I tried the patch but it did not help with ENTER not working, nor the assert. > Does talkback trigger on the nightly crash? No.

Can you get a stack trace leading to this assertion? ###!!! ASSERTION: wrong thread: 'PR_GetCurrentThread() == gSocketThread', file nsHttpConnection.cpp, line 789

Yeah, that's a very troubling assertion. Running that code on the UI thread could result in a crash. Fixing that assertion at least requires finding a way to proxy the GetInterface call back to the socket transport thread. Doing that right probably means fixing bug 289227 and bug 243591. -> Core: Networking

(bug 288729 is for enter in gtk2 builds)

I can now login to my bank again, WFM 2005-04-12-05 Linux (GTK1). Haven't checked the assertion in a debug build yet though.

The assertion is still there, Linux GTK2.

My patch for bug 304286 actually fixed this. The sense of the assertion was simply wrong.