[uClibc] where did the wrappers go?

Carl Miller chaz at energoncube.net
Tue Dec 23 23:34:10 UTC 2003


> > - "ld --nostdinc" apparently can't be trusted to only look where you tell 
> >    it for libraries.
> > 
> > - gcc3 generates code which requires libgcc, some of which requires
> >   glibc symbols and structures (unless we build a uClibc specific libgcc).
> > 
> > Have I got that correct?
> > 
> > I'm curious, as one *should* be able to build and use a library without 
> > building its own toolchain. I'm also in the process of trying to update 
> > Steven Hill's SRPMS, so I want to understand the issues.
> > 
> > Let's start with my first question: why do we need binutils specially 
> > built for uclibc? (I can't see anything different in Mr Hill's binutils, 
> > other than the names of the executables).
> 
> I've found a few relevant threads on the crossgcc mailing list - seems 
> I'm not the only one wishing to simplify toochain building:
> 
> http://sources.redhat.com/ml/crossgcc/2003-11/msg00135.html
> http://sources.redhat.com/ml/crossgcc/2003-11/msg00140.html

And don't forget:
http://sources.redhat.com/ml/crossgcc/2003-12/msg00144.html
http://sources.redhat.com/ml/crossgcc/2003-12/msg00150.html

(that same set of patches, minus the ones to crosstool, were also posted
 to this list at roughly the same time)

Heh-heh, yes, trying to simplify toolchain building, though by making
binutils, gcc, and crosstool specifically recognize uClibc, rather than
by making library-agnostic tools.  That's a far more ambitious goal, I
think.  My hat's off to you if you can pull it off.  Though that is a
good point -- the binutils patches I posted don't actually change anything
other than target tuple recognition.  The generated binutils, once built,
should behave exactly the same whether configured for *-*-linux-uclibc or
*-*-linux-gnu.

Oh, and BTW, Denis Dowling applied my patch, tried building for
i386-pc-linux-uclibc, and found that it didn't work.  It turns out my
patch to recognize target tuples with uclibc didn't cover all the places
where the configure system checks the target.  I just got lucky with my
test case of powerpc, which happened to work.  I'm working on an updated
patch to address this.  Hopefully you'll see it today or tomorrow.


                            ------Carl



More information about the uClibc mailing list