uClibc buildlog for 20060919 snapshot (svn r16143)

[Peter Kjellerstedt]

> -----Original Message-----
> From: Mike Frysinger [mailto:vapier at gentoo.org] 
> Sent: Thursday, September 21, 2006 03:09
> To: Peter Kjellerstedt
> Cc: uClibc
> Subject: Re: uClibc buildlog for 20060919 snapshot (svn r16143)
> 
> On Wednesday 20 September 2006 03:59, Peter Kjellerstedt wrote:
> > >    cris |    cris | Compile | FAIL!
> >
> > Ok, this is weird.  I find the following message in the log:
> >
> > libc/sysdeps/linux/cris/sysdep.S: Assembler messages:
> > libc/sysdeps/linux/cris/sysdep.S:29: Error: symbol `errno' 
> > is already defined
> >
> > But I cannot see how that can happen.  I manually traced all
> > includes from sysdep.S and none of them leads to a file which
> > even references errno...
> 
> the C_SYMBOL_NAME(errno) is expanding to just 'errno' ... the 
> .s looks like:
> 
>  .globl errno
>  .type errno, at object
>  .lcomm errno,4
> 
>  .weak errno
>  errno = _errno
> 
> and we can quickly see where the trouble is :)

Not exactly "quickly", but as far as I can tell the code above fills
no function, and the library built just fine without it (even when 
I disabled thread support), so I just removed all of it...

> > And moreover it does not occur when 
> > I build the latest source here.
> 
> i just updated most of my toolchains to binutils-2.17 and 
> gcc-4.1.1 from gcc-3.4.[56] ... also, i'm still using vanilla 
> gcc rather than the "cris gcc" ... but here it shouldnt matter
> 
> > The log is a little terse, so it is hard to see what has actually
> > happened.  And I am somewhat confused about the fact that it
> > seems like libc is built _after_ libpthread...
> 
> it's built in parallel, so order is arbitrary of subobjects 
> so long as the final shared libs get linked in order

Ah, that explains that. :)

> -mike

//Peter