55567 - Crash or hang when doing force-quit during XPInstall trigger

Status: RESOLVED WORKSFORME

(Core Graveyard :: Installer: XPInstall Engine, defect, P3)

Description

[David Epstein]

Mac build 2000-10-06-08-MN6. Other branch builds as well. This one's hard to pin 
down, and had trouble force-quitting outside of the XPInstall context, but this 
is how to try reproducing:
1. Launch the browser. (preferably after a clean install).
2. Go to http://jimbob/trigger3.html.
3. Select f_addfile_bigfile from Functional test menu. Alternatively, enter a 
bogus URL in the URL field.
4. Press Trigger. Then OK.
5. While the progress bar is going, do a force quit: simultaneously press 
Command (Apple key) - option - esc.
Result: Crash or hang, depending upon what was done before. No Talkback or 
Macsbug for the crash (it was a hard crash).

Comment 1

[David Epstein]

nominating for rtm. It's Mac only, but it leaves things in a nasty state! If 
this bug is not XP related, please reassign to the appropriate component. 

Comment 2

[Daniel Veditz [:dveditz]]

CC'ing Samir for mac expertise, but at this point in the RTM cycle the solution
is "don't do that".

Comment 3

[Samir Gehani]

Force quitting a regular mozilla build while, say, hanging out on a web page,
appears to be clean.  However, force quitting when we are in the middle of an
install (which means probably in the middle of hitting the filesystem) may be
the reason why we are unable to quit cleanly.  I'm not sure there is any easy
remedy to this or if we even need to remedy this.

My understanding from Jimmy is that previously it would simply crash into the
debugger (if you had one installed) but recently it started to crash hard and
you now need to reboot to recover.  I believe that crashing in the debugger
could potentially leave your memory in an inconsistent state and running apps
after coming out of a crash in the debugger with rebooting may anyways possibly
be an invalid state so both the situations warrant a reboot.

Comment 4

[Samir Gehani]

Oops, second last line in my previous comment should read:
after coming out of a crash in the debugger *without* rebooting may anyways
                                             ^^^^^^^

Comment 5

[Jimmy Lee]

Updating QA Contact.

Comment 6

[Brant Gurganus]

By the definitions on <http://bugzilla.mozilla.org/bug_status.html#severity> and
<http://bugzilla.mozilla.org/enter_bug.cgi?format=guided>, crashing and dataloss
bugs are of critical or possibly higher severity.  Only changing open bugs to
minimize unnecessary spam.  Keywords to trigger this would be crash, topcrash,
topcrash+, zt4newcrash, dataloss.

Comment 7

[Chris Casciano]

no crash or hangs or corrumption on multiple attempts to reproduce with various
XPI packages - with OS X 2003042208

WFM?

Comment 8

[Chris Casciano]

resolving WFM since there were no objections