Re: check mips abi x linker emulation compatibility

[Alexandre Oliva <aoliva at redhat dot com>]

On May  7, 2003, Richard Sandiford <rsandifo@redhat.com> wrote:

> I object to this.  One of the main motivations for the test was to
> make sure that NO_ABI behaves correctly.  Adding ABI options to every
> line means we no longer do that.

I see.

> I thought from your message that you were going to just disable the
> combinations that don't work for mips64-linux-gnu.

It turned out that they hadn't worked because of a goof in my patch
for the linker testsuite default.exp.  It's not like the `-32' wasn't
there in my earlier patch, it was just less explicit.

> How about mirroring the ABI selection logic that you added to
> the configure scripts?  Then use that to decide which combinations
> should be tested.

This sounds reasonable, but another alternative that just occurred to
me is to introduce support for -mabi=NO_ABI, that would reset the ABI
to the generic default, such that other forms of ABI inference would
work.  I can't say I like this approach (maybe with a better spelling
for NO_ABI?), but it would enable us to test the ABI-inference even on
targets that do have a default ABI.  Comments?

-- 
Alexandre Oliva   Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer                 aoliva@{redhat.com, gcc.gnu.org}
CS PhD student at IC-Unicamp        oliva@{lsd.ic.unicamp.br, gnu.org}
Free Software Evangelist                Professional serial bug killer