Re: ppc-linux gcc configury patch

[Franz Sirl <Franz.Sirl-kernel@lauterbach.com>]

At 02:43 08.06.99 , Richard Henderson wrote:
> > If there's no compatibility problem with glibc-1.99, changing
> > TEXT_START_ADDR for elf32ppc would be enough, the values in
> > there are already Linux-specific.
>
>Well, not quite.
>
>Mon Jun 16 19:18:21 1997  Geoff Keating  <geoffk@ozemail.com.au>
>
>         * emulparams/elf32ppc.sh (TEXT_START_ADDR): The ABI says `A
>         program base of 0x02000000 is recommended...' because otherwise
>         shared libraries are less efficient.
>
>As such there may well be other elf32ppc users (embedded systems
>folks mostly, so I don't know how much it matters in practice)
>that would prefer to have the default remain unchanged.

Uhm, you left out the most important part of this ChangeLog entry, it 
continues as follows:

         We use 0x01800000 because otherwise it's impossible to branch to
         location 0, for instance if you have an undefined weak symbol.

So it _is_ already Linux/PPC specific.

>Given that I do not expect a native Linux system to have more than
>one of elf32ppc or elf32ppclinux installed, I don't see what
>relevance your LIB_PATH argument is.

See Dan's bugreport on binutils, that's exactly what happens. Without 
special measurements it's impossible to bootstrap gcc-2.95 now. With the 
old binutils the stage compilers will barf, with the new binutils 
egcs-1.1.2 will barf :-(. Wouldn't it be better to split out an 
elf32ppceabi if an embedded guy ever requests it? I would assume embedded 
doesn't need a native library search patch anyway.

Franz.