Re: failure to build base

[Riccardo Mottola]

Hi Niels,

I remember Nicola or Richard already tweaked around those atomic 
operations. I hacked on the assembler of some of those or even 
implemented them... I think they were just enabled with a certain gcc 
version, didn't they?


The platform I am having trouble on is FreeBSD x86-32  gcc 4.2.1

I will test on other platforms.

Riccardo

Niels Grewe wrote:
> Hello Riccardo,
>
> On Sat, May 28, 2011 at 12:10:16AM +0200, Riccardo Mottola wrote:
>    
> > Hi,
> >
> > after the  latest changes and after reconfiguring and cleaning:
> >
> >   Linking tool autogsdoc ...
> > ../Source/./obj/libgnustep-base.so: undefined reference to
> > `__sync_add_and_fetch_4'
> > ../Source/./obj/libgnustep-base.so: undefined reference to
> > `__sync_sub_and_fetch_4'
> > gmake[4]: *** [obj/autogsdoc] Error 1
> >
> >
> > this appears to be related to atomic operations?
> >      
> It is. I introduced a autoconf check that is supposed to avoid just this
> situation. So apparently it's not entirely reliable. With what
> Machine/OS/Compiler combination are you seeing this problem? Also it
> would probably be useful to see the part of config.log related to atomic
> ops.
> Unfortunately, I don't have much time this weekend to fix stuff. Should
> I just roll back the change for the time being?
>
> Cheers,
>
> Niels
>