Updating top-level autoconf to 2.59

[Alexandre Oliva]

On Feb  9, 2007, Ian Lance Taylor <iant@google.com> wrote:

> Alexandre Oliva <aoliva@redhat.com> writes:
>> On Feb  8, 2007, Michael Eager <eager@eagercon.com> wrote:
>> 
>> > I would much prefer explicitly specifying that the build is cross or
>> > native.  I want to specify --cross or --native (or the equivalent).
>> 
>> But then, what if you specify --build=X --host=Y --native (with X!=Y)?

> I don't see a problem with that.  That just means you are building a
> native compiler with a cross-compiler.

Err...  Not really.  autoconf almost doesn't care whether what you're
building is a cross-toolchain component or not.  What matters more to
it is whether the build is native or cross.

> The problematic case is --host=X --target=Y --native, with X != Y.
> For that case, the configure script should simply give an error.

> Which may just mean that the --native and --cross options are
> themselves not well defined.

Exactly.  So why not just key off the presence of --host (or --target,
for that matter)?

> There is no obvious reason why I can't specify the host when not
> building with a cross compiler.

You can specify host in this case.  And the build ought to work just
fine.  It's just that you'll trigger cross-compilation paths, which is
probably good for testing actual cross compilation.

> From my perspective it's just an odd autoconf rule.

It is, indeed.  IIRC I resisted it and was outvoted.  But I didn't
resist it because it didn't make sense, but rather because I knew it
would break build scripts.

-- 
Alexandre Oliva         http://www.lsd.ic.unicamp.br/~oliva/
FSF Latin America Board Member         http://www.fsfla.org/
Red Hat Compiler Engineer   aoliva@{redhat.com, gcc.gnu.org}
Free Software Evangelist  oliva@{lsd.ic.unicamp.br, gnu.org}