[uClibc] frv-uclinux port updates

[Joakim Tjernlund]

> On Sun Dec 05, 2004 at 12:06:54AM -0200, Alexandre Oliva wrote:
> > Here's an updated version of the patch I posted about a month ago.  It
> > leaves -nostdinc alone, and uses -print-file-name=include instead of
> > -print-search-dirs to figure out where GCC's internal headers are.
> > 
> > Please let me know whether there are any portions of this patch you'd
> > like me to break into smaller pieces, to rework, or to give up trying
> > to get into uClibc :-)  Thanks,
> 
> All the non-generic stuff is in (except where you removed
> -nostdinc anyway even though you said you didn't ;-).  For the
> generic stuff, it makes things much more ugly with all those
> unsighly macros so you are going to need to spoon feed it to me
> one bit at a time.  For example your _dl_malloc() and _dl_free()
> stuff looks to be quite independent of everything else.  Your
> _dl_find_hash() changes look to be seperable.  For the rest,
> perhaps you could feed things to me one or a few macros at a time
> so the pieces will be in smaller more logically cohesive chunks.
> The whole lot in one patch is too much to choke down as a single
> blob,

There are a lot of code that does 
  loadaddr + tpnt->dynamic_info[DT_X];
Most of these calculations can be done once in __dl_parse_dynamic_info(). Glibc
already does this.

That will save space & speed and hopefully will reduce a few of the macros introduced
by FRV.

Not sure I like what you have done to dl_find_hash(). I think it would be better if you
just changed dl_find_hash() to return a elf_resolve * instead.

I hope that you will find the time to split up the patch in smaller chunks, as this will
make it a lot easier to comment on.

Sidenote: I think you need to use the mytpnt arg to dl_find_hash in elfinterp.c, otherwise
you may have problems with dlopened libs with RTLD_LOCAL scope.

 Jocke