FYI: Support -r* processor type option for SGI C++ compiler

[Ralf Wildenhues]

* Albert Chin wrote on Sat, Jan 22, 2005 at 08:27:57PM CET:
> On Wed, Dec 08, 2004 at 06:42:49PM +0100, Ralf Wildenhues wrote:
> > * Albert Chin wrote on Wed, Dec 08, 2004 at 05:04:19AM CET:
> > > With the SGI C++ compiler, -r* selects the CPU type:
> > >      -rprocessor Specifies the code scheduler.  processor can be one of 
> > > the
> > >                  following options:
> > > 
> > >                  Option  Action
> > >                  4000    Schedules code for the R4000 processor.
*snip*
> > 
> > This:
> > 
> > >                  This option adds the following to the head of the library
> > >                  search path, where processor is as you specified:
> > > 
> > >                       -L/usr/lib{32,64}/mips{3,4}/processor
> > 
> > sounds to me like we may need to know more about this option and not
> > just allow it through plainly.  What kind of libraries do these paths
> > usually contain (e.g., low-level stuff like libc only as compared to
> > general-purpose libraries)?  Does the option also set rpath or have
> > other side-effects?  Are these options really like -mips[0-9] except
> > they allow finer-grained selection (then your patch would probably be
> > ok)?
> 
> Yes, they allow finer-grained selection:
*snip*
>   $ cc -v a.c
>   ...
>   /usr/lib32/cmplrs/ld32 -call_shared -no_unresolved -transitive_link
>   -elf -_SYSTYPE_SVR4 -show -mips3 -n32 -L/usr/lib32/mips3/r4000
>   -L/usr/lib32/mips3 -L/usr/lib32 /usr/lib32/mips3/crt1.o a.o
>   -dont_warn_unused -Bdynamic -lc /usr/lib32/mips3/crtn.o -warn_unused 
> 
>   $ cc -r5000 -v a.c
>   ...
>   /usr/lib32/cmplrs/ld32 -call_shared -no_unresolved -transitive_link
>   -elf -_SYSTYPE_SVR4 -show -mips3 -n32 -L/usr/lib32/mips3/r5000
>   -L/usr/lib32/mips3 -L/usr/lib32 /usr/lib32/mips3/crt1.o a.o
>   -dont_warn_unused -Bdynamic -lc /usr/lib32/mips3/crtn.o -warn_unused 

So I gather it's not generally possible to mix these, and thus, you'll
have to specify these at configure time, and libtool will find
predep_objects and postdep_opjects accordingly.

Guess we can let it through then.

I've applied your patch to all three branches.

Regards,
Ralf