[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#624354: Processed: Re: Bug#624354: ./xpcshell: error while loading shared libraries: ./libxul.so: R_PPC_REL24 relocation at 0x0f9f0148 for symbol `_restgpr_29_x' out of range



On Wed, Aug 31, 2011 at 08:39:05AM +0200, Michel Dänzer wrote:
> On Mit, 2011-08-31 at 07:35 +0200, Mike Hommey wrote: 
> > On Wed, Aug 31, 2011 at 07:26:10AM +0200, Mike Hommey wrote:
> > > On Wed, Aug 31, 2011 at 07:05:04AM +0200, Mike Hommey wrote:
> > > > On Wed, Aug 31, 2011 at 06:35:09AM +0200, Mike Hommey wrote:
> > > > > Interestingly, according to bug #639851, 6.0-2 didn't have the problem.
> > > > > Which suggests something else broke.
> > > > 
> > > > Actually, it must have worked by luck in 6.0-2 for that user, because
> > > > the binary is affected the same way.
> 
> Yeah, 6.0-2 also worked for me originally, but it no longer did after
> downgrading to it from 6.0-4.
> 
> Again, R_PPC_REL24 relocations in shared objects are ticking time bombs,
> which may happen to work most of the time... but they're always a bug.
> 
> 
> > > > So, it would be interesting to find what particular library is exporting
> > > > the _rest* symbols this time, since libstartup-notification doesn't
> > > > anymore.
> > > 
> > > And the winner is libevent.
> > 
> > iceweasel built against 1.4.13-stable-1 which was built a long time ago,
> > like libstartup-notification had, and 1.4.14b-stable-1 currently in
> > unstable doesn't export the symbols. Same conclusion as with
> > libstartup-notification: old toolchain bug.
> > 
> > Still, I think it's wrong that a random library can highjack symbols
> > that gcc uses to call its own stuff from libgcc. Are there actual
> > legitimate uses of that "feature"?
> 
> I don't know, and I agree this seems like a toolchain bug. But why not
> build with -O2 and stay out of the mess?

I'm not saying that I don't want to fix the mess on iceweasel end. I'm
saying that even if we fix the mess on iceweasel end, there is still a
toolchain problem.

On iceweasel end, however, I'm pondering checking the resulting binaries
instead of changing the build options.

Mike



Reply to: