285743 - after running under solaris 5.9, using the same profile under linux causes thunderbird to exit

Status: RESOLVED INCOMPLETE

(Thunderbird :: General, defect)

Description

[Todd Blake]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.7.6) Gecko/20050225 Firefox/1.0.1
Build Identifier: 

I'm currently sharing user directories between solaris 5.9 machines and linux
machines running redhat 9, and Scientific Linux 4.1alpha(a free clone of Redhat
Enterprise Linux)

If I have no profiles on my user account and launch thunderbird under linux, a
profile is created and thunderbird runs fine.  If I do the same under solaris,
everything runs fine.  Solaris reads the profiles created under linux just fine
as well.

The problem occurs when a profile is used beneath thunderbird under solaris. 
Once this is done, any attempts to use the same profile under linux will cause
thunderbird to exist with a status code of 1.

Also, if multiple profiles exist, it will get far enough to let you pick one, so
picking a profile solaris has never touched will let it execute under linux.  If
you pick a profile that was also picked under solaris, it will crash.

Reproducible: Always

Steps to Reproduce:
1. Run thunderbird under linux, create a profile, then manually exit
2. Run thunderbird using the same user account and directories(shared via NFS)
under solaris 5.9, then manually exit
3. Try to run it again under linux, it will fail.

Actual Results:  
Thunderbird was executed from the command line, and the system simply returned
to the shell prompt.

Expected Results:  
Started thunderbird

To get a good strace and truss output, I ran thunderbird as such:

1. Thunderbird was unpacked(not installed) into /usr/site/thunderbird-1.0
2. Thunderbird was launched with these commands under bash to execute the
binary, not the shell script that eventually launches the binary.
$ export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/site/thunderbird
$ /usr/site/thunderbird/thunderbird-bin 

Attaching strace output from a working profile under linux and a crashed profile
under linux.  Also attaching truss output from a succesful run under solaris.

Comment 1

[Todd Blake]

Hrm, the outputs from strace and truss are too big to attach.  I'm putting them
on  the web, I'll post the URL momentarily.

Comment 2

[Todd Blake]

(In reply to comment #1)
> Hrm, the outputs from strace and truss are too big to attach.  I'm putting them
> on  the web, I'll post the URL momentarily.

Here's the URL to three files
1) strace output from thunderbird crashing under linux using a profile that
solaris&linux touched
http://tiger.towson.edu/~tblake1/thunderbirdbug/thunderbird_crashes_linux

2) strace output from thunderbird working under linux using a profile that only
linux touched
http://tiger.towson.edu/~tblake1/thunderbirdbug/thunderbird_works_linux

3) truss output from thunderbird working under solaris using a profile that
solaris&linux touched
http://tiger.towson.edu/~tblake1/thunderbirdbug/thunderbird_works_solaris

Comment 3

[Boris Zbarsky [:bzbarsky]]

Todd, does talkback trigger?  (Based on that truss output, doesn't look like it
would...)

This sounds like an endianness issue somewhere.  In the "working" output we
stat/open/read xpti.dat twice, then start calling stat on the xpt files.  In the
broken output, we stat/open/read it once, then get a brk(), do a pipe() call,
mess with fcntl64, try a read(), get back a -1 return, close filehandles, unlink
the lock file, and call exit_group.

Doesn't sound like very good startup behavior to me (no crash, silent
_controlled_ exit, but no indication of what's wrong to the user).

Comment 4

[Todd Blake]

Nope no talkback trigger, just a return to the command prompt.  I kinda figured
it was an endianness issue, and I'm kinda figured I'll just end up SOL since
there's no easy way for thunderbird coders to get around this one.

Comment 5

[Boris Zbarsky [:bzbarsky]]

> there's no easy way for thunderbird coders to get around this one.

Sure there is.  They know the endianness of the platform they're running on and
all.  These issues have come up before, and have always been due to bad
assumptions on the part of the app...

Comment 6

[Todd Blake]

This is why I'm not a programmer :-).

Comment 7

[Wayne Mery (:wsmwk)]

Todd. If you still see this problem, can you try TB3 alpha 2 (just out)? 
 http://www.mozillamessaging.com/en-US/thunderbird/early_releases/
Perhaps it will get us a crash dump.  

Bug 331511 is also xplatform issue.

Comment 8

[Wayne Mery (:wsmwk)]

Emily, are  you aware of any problems related to reporter's Steps to Reproduce?

OP writes "I don't have the infrastructure to test".

Comment 9

[Wayne Mery (:wsmwk)]

no response to ludo's request for help http://groups.google.com/group/comp.unix.solaris/browse_thread/thread/288ce4f88378fd25?pli=1