[uClibc] RE: [uClibc-cvs] svn commit: trunk/uClibc/ldso/ldso: arm cris i386 m68k mips powerpc sh sh64 etc...

[Joakim Tjernlund]

> On Tue, Mar 29, 2005 at 03:13:34PM +0200, Joakim Tjernlund wrote:
> > > At least MIPS and sh64 need to get at the symtab for their relative
> > > relocation handling. These last few ldso changes just seem to drop
> > > relative relocations on MIPS entirely, so that's a problem there as well
> > > (at least adding back in R_MIPS_REL32 should be a walk in the park).
> > 
> > Didn't know that there were RELATIVE relocs on MIPS. Glibc doesn't think so.
> > 
> There are, but IIRC they are rather expensive and tend to be avoided in
> practice. I haven't seen the glibc source, but it's possible that this is
> the reason for it.
> 
> On the other hand, it's been years since I last touched MIPS, someone can
> feel free to correct me if I am confused on this matter.
> 
> > > Until this is fixed, _dl_fixup() and things of that nature will pretty
> > > much just die in horrible and interesting ways for relative relocations
> > > on these platforms.
> > 
> > I don't think MIPS is broken, nobody has complained about that.
> > 
> Nobody has complained because the relocation types that would be broken
> probably don't show up very often in the wild (note the _relative_
> keyword in the preceeding paragraph) for MIPS. At least not often enough
> for the few people testing to notice. It's more of a corner case I guess.

I see, thanks.

> 
> > Should I commit http://cvs.uclibc.org/lists/uclibc/2005-March/011465.html first?
> > 
> Yes, that would make things easier.

Done.