OpenOffice compilation Failures, LFS Unstable

Matthew Burgess matthew at linuxfromscratch.org
Wed Dec 29 01:45:58 PST 2004


DJ Lucas wrote:
> On Mon, 2004-12-27 at 15:48 -0700, Nathan Coulson wrote:
> 
> 
>>>BTW, I think I mentioned this was LFS unstable in my origional email
>>>
>>>binutils 2.15.94.0.1
>>>gcc 3.3.4 (compiled to /opt)
>>>and glibc 1122
>>>compiling openoffice 1.1.4 [got the same thing compiling 1.1.2]
>>
>>Reading up on what I have seen, my guess would be that fedora added
>>extra checks to glibc [aka, guarding against calling free twice on the
>>same pointer].
>>
> 
> 
> Actually, I went forward and installed it this way from my last build in
> unstable, and it seems to work...except for no help.  Honestly, no idea
> if it's related, there is no telling on my unstable system.  I couldn't
> remember exactly which directory the error occured in.  Question, is
> that a warning, a failure or what?  No magic words "Error" or "Warning"
> there.
> 
> *** glibc detected *** double free or corruption (!prev): 0x08069288 ***

Jim Gifford's seen something similar IIRC (a shell script appeared to be 
causing bash to output something similar).  Based on the fact that the 
application doesn't terminate, I'd treat that output as a warning, but 
only because it's not glibcs place to be be aborting an application.

For double free bugs like this, it might be worth running the affected 
binary through 'valgrind' to see where the problem actually lies. 
Having said that running a debug version of OpenOffice would probably be 
slow enough - running that through valgrind is going to be painful, to 
say the least, I would imagine!

Cheers,

Matt.



More information about the blfs-dev mailing list