31858 - RegXPCOM crashes on Mac (beta1_BRANCH)

Status: VERIFIED FIXED

(Core :: XPCOM, defect, P3)

Description

[J.J. Enser]

RegXPCOM has become very unstable, crashing _most_ of the time. This prevents 
BETA1 daily verif. build to complete and delays delivery to QA.

I tried to find a reproducile scenario, but it's not easy. App sometimes runs 
without crashing, sometimes crashing right on startup, creating a 0K length 
Component Registry, sometimes crashing before quitting, creating a full length (+
~320K) registry.
this happnes with a memory partition bumped up to 24 MB.

I intended to attach a Macsbug log, but 'stdlog' fails because 'the sile system 
is busy'. :-(

Help!

Comment 1

[J.J. Enser]

nominating for beta 1

Comment 2

[leger]

Putting on PDT+ radar for beta1.  Sending over to sfraser...can you help?

Comment 3

[Suresh Duddi (gone)]

Robert, please help simon if he needs with a stack trace if you can reproduce 
it. 

Comment 4

[Simon Fraser [no longer active]]

A stack isn't gonna help here; the one crash I saw was somewhere in the Toolbox, 
another in some other random location. The heap wasn't corrupted, so I've no idea 
what's happening. My only hunch is that this app runs (in an optimized build) 
without a window, so if we have any code that gets hit that assumes a window 
(like in widget or gfx) then bad things might happen.

Comment 5

[Suresh Duddi (gone)]

Here is one way to find out: remove sets of dlls from components/ , delete 
component.reg and run regxpcom until we dont crash. That way we can narrow down 
the offending dll.

Would that work ?

Comment 6

[Simon Fraser [no longer active]]

Well, I did an optimized, branch build on another machine, ran RegXPCOM 20 times, 
and it didn't crash once. I'm starting to suspect the machine.

Comment 7

[Simon Fraser [no longer active]]

jj and i tried running RegXPCOM in various ways on the release build machine, and 
still could not get it to crash. We suspect some deeper memory corruption/leakage 
by the other tools to be the cause of this.

Comment 8

[J.J. Enser]

fyi, there is a beta1 respin in progress right now, and it also "passed" RegXPCOM 

without crashing. This was on a diffferent machine than the one which usually 

crashes... Defenitely smells like a config/OS/setup issue.



Will try to mirror the 'trunk' verif. build mac on the 'branch' verif. build mac 

and see if it's happier tomorrow morning. 

Comment 9

[Jim Roskind]

This is sure souinding like a worksforme bug.
If we can't repro it tomorrow, lets get it off the radar and mark it as such.
Thanks,
Jim

Comment 10

[Simon Fraser [no longer active]]

Reassign to jj to figure out if this is machine juju

Comment 11

[J.J. Enser]

removing PDT+ to get it off the beta1 radar, since it works fine on a different 
build machine.
Will update status if problem reappears in the near future

Comment 12

[leger]

Putting on the PDT- radar for beta1.

Comment 13

[J.J. Enser]

upgraded build mac to OS 9 and reinstalled build environment. no longer crashes.

Comment 14

[leger]

Adding crash keyword

Comment 15

[Tracy Walker [:tracy]]

- Per last comments, age of bug, and no reopen - Marking Verified/Fixed.  Please 
reopen if still a problem.