> -----Original Message-----
> From: Akim Demaille [mailto:[EMAIL PROTECTED]]
> Sent: Tuesday, May 23, 2000 1:39 PM
> To: Alexandre Oliva
> Cc: Mo DeJong; [EMAIL PROTECTED]; [EMAIL PROTECTED];
> [EMAIL PROTECTED]
> Subject: Re: automatic cross compiler search patch
>
>
> >>>>> "Alexandre" == Alexandre Oliva <[EMAIL PROTECTED]> writes:
>
> Alexandre> On May 21, 2000, DEMAILLE Akim <[EMAIL PROTECTED]> wrote:
> >> This point is that if the user did not --build, then
> >> $build_alias='' likewise for the two others. Before, $build_alias
> >> would have been the same as as $build, i.e. computed by
> >> config.guess.
>
> Alexandre> I fear backward compatibility problems. What if
> Alexandre> configure.in refers to $build, $host or $target before
> Alexandre> calling AC_CANONICAL_*? Other than this consideration, the
> Alexandre> patch seems fine.
>
> So, do you people think the patch should be applied as is? Anyway, if
> there are real problems, we will quickly know during the beta testing.
>
I just read this thread now, so was not able to react earlier, but this
patch is just correcting something that bothered me a lot (and cause some
bad head-scratching :-(): I do not like variables describing a given
configuration to change value during the script.
It's perfectly fine to have the value defined after some AC_macro is called,
and undefined before, but having the variable defined and suddenly change
value to some variation usually having the same meaning is annoying to say
the least :-)
So I strongly vote for this patch.
This whole thing about setting host from --host, then using config.sub to
compute a new value for host while saving the old value in host_alias seems
complicated and a bit kludgy; it seems it was another historical complexity
that appeared when config.* was created.
What is even worst is that if you need the value given by the user (in the
"old" scheme) you have to test host before AC_CANON and host_alias after; so
if I start without the need for canonicalized names, I will test host then
adding the AC_CANON call later on will break my configure script as now I
have to test fro build_alias...
Now that we redefine cleanly what cross-compilation mean (--host specified)
it seems the right time to define cleanly how we can get the value of the
corresponding option as passed by the user, and how to get the canonicalized
one if needed, both in an uniform way.
Regards,
Bernard
--------------------------------------------
Bernard Dautrevaux
Microprocess Ingéniérie
97 bis, rue de Colombes
92400 COURBEVOIE
FRANCE
Tel: +33 (0) 1 47 68 80 80
Fax: +33 (0) 1 47 88 97 85
e-mail: [EMAIL PROTECTED]
[EMAIL PROTECTED]
--------------------------------------------
