committed: merge with libada-branch

[Joel Sherrill]

Daniel Jacobowitz wrote:

> On Tue, Feb 10, 2004 at 12:05:17PM -0800, Zack Weinberg wrote:
> 
>>Daniel Jacobowitz <drow@mvista.com> writes:
>>
>>
>>>On Tue, Feb 10, 2004 at 11:56:17AM -0800, Zack Weinberg wrote:
>>>
>>>>The set of multilibs is determined entirely by the target
>>>>architecture.  Building a cross compiler to just about any
>>>>embedded target (mips-elf for example) will involve multilibs.
>>>
>>>You can also build them for an i386 native if you prefer.  GCC will
>>>happily build multilibs for anything you choose to specify; you could
>>>probably multilib on -g if you wanted to.
>>
>>I was trying for a suggestion that didn't involve mucking with
>>t-fragments, or have I missed something?
> 
> 
> Nope.  The mucking is pretty straightforward though.
> 
> MULTILIB_OPTIONS = g
> MULTILIB_DIRNAMES = g
> 
> is probably all that is necessary.  Not even sure about the second
> line.
> 

If you want to play on the i386 and get multilibs, 
config/i386/t-rtems-i386 is set up to build a number of
them.  We end up with these library variants:

athlon/libc.a
k6/libc.a
m486/soft-float/libc.a
m486/libc.a
mpentium/libc.a
mpentiumpro/libc.a
soft-float/nofp/libc.a
soft-float/libc.a
libc.a

Just cut those and put them in another t- file.

Be warned, turning on multilibs can trigger bugs which were
overlooked before.  I recall that it took a few rounds of bug reports
before we could reliably build i386-rtems with all those multilibs.

--joel