Re: AC_CYGWIN etc. (Was: AC_OBJEXT again)

[Akim Demaille]

>>>>> "Alexandre" == Alexandre Oliva <address@hidden> writes:

Alexandre> On Dec 14, 2000, Akim Demaille <address@hidden> wrote:
>> I plan to have the two tests: the a.exe one (in a conftestdir as
>> was to be suggested by Alexandre), and the -o conftest$ac_exeext
>> one.

Alexandre> To be sincere, I hadn't thought of this option.  But, now
Alexandre> that you mentioned, I see it may introduce other (unlikely,
Alexandre> I agree) problems in itself.  Think of CC, CPPFLAGS,
Alexandre> CFLAGS, LDFLAGS or LIBS containing some options that
Alexandre> reference relative filenames.  If you enter conftestdir,
Alexandre> these options may break the compilation/linking.  Ugh!

Very nice point, but after some more thought I don't think it matters
at all.  I mean this is very very early in the process, there is
nothing in confdefs.h, and since configure is being run, we are
probably not using anything from the package itself.  So remain only
absolute parameters (installed headers, libs etc) which should not be
affected by conftestdir.

But maybe I'm not really considering the real problems, I trust you.

Then, frankly, using a.out and a.exe doesn't sound that bad to me.  If
you want we can save and restore them.

>> As for the other cross compilers, let them receive our signal: we
>> are setting a standard they have to follow :)

Alexandre> This doesn't fit in very well with autoconf's standards :-)

Huh?  Then there might be something points I didn't catch properly :)