943463 - B2G desktop startup crash | libxul.so!MarkInternal<JSObject>

Status: RESOLVED DUPLICATE of bug 942421

(Core :: JavaScript Engine, defect)

Description

[Andrew Halberstadt [:ahal]]

Attachment: crash stack

I'm encountering a crash on startup for b2g desktop. I've attached the crash stack and the mozconfig I used.

STR:
* build with attached mozconfig
* <objdir>/dist/bin/b2g -profile path/to/gaia/profile

Results:
I see the locale picker very briefly and then the attached crash occurs.

Additional info:
* First noticed this yesterday (regression range unclear)
* I'm using mozilla-central with gaia master, both recently pulled.
* I've tried building gaia with many different permutations (DESKTOP_SHIMS, NOFTU, DEBUG) and they don't seem to make any difference.
* I'm on Fedora 19 64 bit

Comment 1

[Andrew Halberstadt [:ahal]]

Attachment: mozconfig

Comment 2

[Andrew Halberstadt [:ahal]]

I can reproduce this on my laptop which is also running Fedora 19 64bit.

Comment 3

[(no longer active)]

Can you please bisect this?  Thanks!

Comment 4

[Andrew Halberstadt [:ahal]]

Should have checked this earlier, but I can't reproduce on downloaded builds from ftp.m.o. I'm not sure what I'm doing differently, my mozconfig is the same as the one in https://developer.mozilla.org/en-US/Firefox_OS/Using_the_B2G_desktop_client#Create_a_mozconfig except for the ENABLE_MARIONETTE=1 at the end.

I'll try building from an old m-c checkout.

Comment 5

[Gregor Wagner [:gwagner]]

Maybe related to https://tbpl.mozilla.org/php/getParsedLog.php?id=31178503&tree=Pine&full=1#error1 ?

Comment 6

[Andrew Halberstadt [:ahal]]

Sorry it's taking so long, needing to build each time has been a bit time consuming. I've narrowed it down to sometime between Nov 21st and Nov 22nd:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=09e33431c543&tochange=b9980eb88a26

I'll continue chipping away at it, I'll start using my laptop as a dedicated build machine so hopefully it'll go a little faster from here on out.

Comment 7

[Andrew Halberstadt [:ahal]]

So the culprit is http://hg.mozilla.org/mozilla-central/rev/3b9e118ded0f

Ehsan, any ideas?

Comment 8

[(no longer active)]

Well crap!

Two questions:

1. Can you please try the latest inbound which has fixes to a couple of regressions from that bug?
2. If 1 doesn't work, can you please open js/src/moz.build, find the UNIFIED_SOURCES chunk here (http://hg.mozilla.org/mozilla-central/rev/3b9e118ded0f#l9.13), and then start "bisecting" that list by moving parts of it to SOURCES until you find the minimum number of files which need to be moved to SOURCES to make this bug go away?

(Also, out of curiosity, what compiler are you using?)

Comment 9

[Andrew Halberstadt [:ahal]]

Yay, it worked on inbound! For the record I was using gcc 4.8.2

Comment 10

[(no longer active)]

Thanks for verifying.  I'm thinking that this is probably a dupe of bug 942421.