1026329 - SeaMonkey >= 2.29 crashes when clicking on URL bar

Status: RESOLVED WORKSFORME

(SeaMonkey :: General, defect)

Description

[kevink9876543]

User Agent: Mozilla/5.0 (X11; Linux i686; rv:33.0) Gecko/20100101 Firefox/33.0 SeaMonkey/2.30a1 (Beta/Release)
Build ID: 20140616135943

Steps to reproduce:

1) Start SeaMonkey, but don't interact with it at all for at least a minute or so.
2) Click on the URL bar (search bar might work too, but not completely sure).

(First noticed this with SeaMonkey 2.29 nightlies, but had trouble finding consistent STR.)


Actual results:

SeaMonkey crashes.  Here are a couple of submitted crash reports:
http://crash-stats.mozilla.com/report/index/bp-9cf05a71-9ab1-4f9a-b230-2e53b2140617
http://crash-stats.mozilla.com/report/index/bp-77dc481d-5618-464a-87f4-8bbbd2140617


Expected results:

SeaMonkey should not have crashed - in fact, nothing special should have happened at all.

Comment 1

[Philip Chee]

You're doing your own builds right? Therefore crash-stats.mozillan doesn't have symbol information for your build. If you build with debug-symbols I think dbg will be able to use them. I'm not using Linux so I don't know what the perquisites are.

Comment 2

[kevink9876543]

Attachment: Test profile where I reproduced this issue

Can't reproduce in a clean profile, but can usually reproduce with this test profile.  The included "Advanced Cookie Manager" addon is my own fork of
https://addons.mozilla.org/firefox/addon/cookie-manager/versions/5.2
by Jayapal Karakala, and the major changes I made are removing the Facebook integration and adding SeaMonkey support.

Comment 3

[therube]

Seeing nothing unusual on my end, on WinXP.

Comment 4

[Philip Chee]

So WAG. Did you try SeaMonkey is safe mode? What do you have in your mozconfig? Did you try building with a minimal mozconfig?

Comment 5

[kevink9876543]

(In reply to Philip Chee from comment #4)
> Did you try SeaMonkey is safe mode?
Just did, with the above test profile... and reproduced the crash.

> What do you have in your mozconfig?
My mozconfig is

ac_add_options --enable-application=suite
mk_add_options MOZ_OBJDIR=@TOPSRCDIR@/objdir-sm-release
mk_add_options MOZ_MAKE_FLAGS="-j4"

Comment 6

[kevink9876543]

Oh, and also of note is that I'm building with the patch from Bug 987464 comment 0, and my build system is an up-to-date Ubuntu 12.04 (32-bit) with no special configuration in this respect.  But other than that, my build process is pretty much routine / as the docs on MDN say.

Comment 7

[Philip Chee]

try --disable-optimize and/or:

export MOZ_DEBUG_SYMBOLS=1
ac_add_options --enable-debug-symbols="-gdwarf-2"

https://developer.mozilla.org/en/docs/Building_Firefox_with_Debug_Symbols
http://shlomif-tech.livejournal.com/62731.html

Comment 8

[kevink9876543]

(In reply to Philip Chee from comment #7)
> try --disable-optimize and/or:
> 
> export MOZ_DEBUG_SYMBOLS=1
> ac_add_options --enable-debug-symbols="-gdwarf-2"
Can't:

creating ./config.status
Traceback (most recent call last):
  File "./config.status", line 992, in <module>
    config_status(**args)
  File "/media/mozilla/comm-central/mozilla/python/mozbuild/mozbuild/config_status.py", line 134, in config_status
    emitter = TreeMetadataEmitter(env)
  File "/media/mozilla/comm-central/mozilla/python/mozbuild/mozbuild/frontend/emitter.py", line 74, in __init__
    mozinfo.find_and_update_from_json(config.topobjdir)
  File "/media/mozilla/comm-central/mozilla/testing/mozbase/mozinfo/mozinfo/mozinfo.py", line 151, in find_and_update_from_json
    build = MozbuildObject.from_environment()
  File "/media/mozilla/comm-central/mozilla/python/mozbuild/mozbuild/base.py", line 201, in from_environment
    raise ObjdirMismatchException(topobjdir, _config_topobjdir)
mozbuild.base.ObjdirMismatchException: Objdir mismatch: /media/mozilla/comm-central/objdir-sm-release/mozilla != /media/mozilla/comm-central/mozilla/objdir-sm-release/mozilla
configure: error: /media/mozilla/comm-central/mozilla/configure failed for mozilla
*** Fix above errors and then restart with               "make -f client.mk build"
make[1]: *** [configure] Error 1
make[1]: Leaving directory `/media/mozilla/comm-central'
make: *** [/media/mozilla/comm-central/objdir-sm-release/Makefile] Error 2


Related to bug 987464?

Comment 9

[kevink9876543]

OK, got the build working by changing my objdir path... but despite ac_add_options --enable-application=suite it built Thunderbird instead of SeaMonkey!
(I'm confused.)

Comment 10

[Philip Chee]

If c-c can't find your mozconfig it builds Thunderbird by default. Here's how I start a build:

$ MOZCONFIG=/c/t1/hg/comm-central/mozconfig-sm mozmake.exe -f client.mk

Comment 11

[Philip Chee]

Attachment: Patch to backout mozilla-central Bug 1035096 and Bug 1035099

I have this backout patch applied to mozilla-central so that I can get successful builds.

Comment 12

[Philip Chee]

Attachment: moz.build patch

(No bug number yet) comm-central patch - build system is looking in the wrong place for application.ini.h

Comment 13

[kevink9876543]

Well, this is weird.  With a debug build*, it's harder to reproduce this, and even harder running it inside gdb (in fact, seems impossible to reproduce if trying to debug using mach).  But I eventually reproduced (something at least similar to) this bug invoking gdb directly, and it gave me this:

Program received signal SIG38, Real-time event 38.
0x08060a5f in mozilla::detail::AddUintptrToHash<4u> (aHash=0, aValue=
    2888903104) at ../dist/include/mozilla/HashFunctions.h:129
129	{
(gdb) continue
Continuing.

Program received signal SIGSEGV, Segmentation fault.
0xa98f0d20 in ?? ()
(gdb) continue
Continuing.


Then SeaMonkey continued happily.
???


* mozconfig:
ac_add_options --enable-application=suite
mk_add_options MOZ_OBJDIR=/media/mozilla/objdir-sm-release
mk_add_options MOZ_MAKE_FLAGS="-j4"

ac_add_options --disable-optimize
export MOZ_DEBUG_SYMBOLS=1
ac_add_options --enable-debug-symbols="-gdwarf-2"
ac_add_options --disable-install-strip

Comment 14

[kevink9876543]

Gecko 32 / SeaMonkey 2.29 is now in beta, and this bug hasn't yet been fixed.  Is there any other information I need to provide?

Comment 15

[kevink9876543]

Still happening in 2.31...
Since it seems that the official SeaMonkey Linux build machines are at least able to build, is there a recent official Linux i686 build (preferably a nightly or aurora) that I could download to try to reproduce this and submit a crash report for which you have a symbols file?

(To note, my self build of 2.28 release done on the same build machine is not affected, so whatever is causing this didn't get backported.)

Comment 16

[kevink9876543]

Official Linux i686 nightlies are back now, and I can't reproduce this bug with official build 20140901003005...
What mozconfig and OS were used to build 20140901003005?

Comment 17

[kevink9876543]

I've switched to a CentOS 6.5 build machine, and I cannot reproduce this in a self build from today's code run straight from the objdir (run on my prior build machine).  I'd like to be able to try a packaged self build to make really sure, but packaging my build doesn't work, but since I think I could reproduce it running straight off the objdir before, marking as WORKSFORME.