[Bob Proulx]
> I disagree.  It is wrong for config.guess to use 'uname -p' unless
> it knows it is an okay thing to do on that particular host since
> that is a non-standard thing to do.  It has no right to count on it
> being there.

Feel free to discuss this with the maintainers of config.guess.


> Perhaps on the configure command line you could force the selection.
> Something like this example.  But these are from an x86.  Replace with
> your appropriate values.
> 
>   ./configure --host=i386-linux --build=i386-linux

Well, of course this works, but it is not an option when I use the
same compile script to compile the program on 9 different
architectures.  I would rather fix the auto-detection.

> Perhaps you could update your config.guess and config.sub on a
> failing package to the newest available and see if that helps.  Both
> of those are needed and go together.  I believe the canonical and
> newest versions are here.  (I hope I have that right.)  On hosts
> that are not as mainstream I frequently find I need to update those
> files.

The problem occurs with the latest config.guess.  It only occurs if
GNU uname is first in the PATH.  If I make sure /usr/bin/uname is
first in the path, the output from config.guess is correct.

I do not understand why you insist there must be something else wrong.
Is it so hard to make GNU uname return 'powerpc' instead of 'unknown'
as the hardware type on Darwin?


_______________________________________________
Bug-sh-utils mailing list
[EMAIL PROTECTED]
http://mail.gnu.org/mailman/listinfo/bug-sh-utils

Reply via email to