Re: more Autotest thoughts

[Ralf Wildenhues]

Hi Stepan,

* Stepan Kasal wrote on Fri, Apr 29, 2005 at 11:38:04AM CEST:
> On Fri, Apr 29, 2005 at 10:15:11AM +0200, Ralf Wildenhues wrote:
> > The only thing the test suite needs to know is: Can we run executables
> > intended for $build at all or not?  This may be found out with
> > AC_RUN_IFELSE for once and all, [...]
> 
> You don't have to use AC_RUN_IFELSE.  Just ``test $cross_compiling = yes''.

No.  This is exactly what I would like to avoid.

> There are cases when cross_compiling = yes, yet you can run the executables,
> for example when you have
>       --build=i686-pc-gnu --host=i386-pc-gnu
> or even
>       --build=i686-pc-gnu --host=i686-gnu

or when you have
- wine installed, or any (other) system simulator,

or, after David Heine's recent patch or similar in place,
- any way to cause an executable to be executed on a $build system
(for example, you may have to temporarily chroot somewhere, or ssh
on the $build system where the nfs volume is mounted as well, or even
scp the binary to a $build system if it's only a simple one without
dependencies, and execute it there).

and I would very much like to be able to use that.  This is the goal.

Hey, come to think of things like xen, the possible ways of doing this
really changed in the last few years.

> But AC_RUN_IFELSE doesn't help you here, since it doesn't even try to run
> the executable when cross_compiling = yes.

Ahh, right. 

> If you wanted to be absolutely sure, you'd have to try to run an executable
> on your own.  But I don't think it's worth it.

Making this easily possible would be a great step forward for Autoconf
IMVHO.  David Heine's patch, and maybe a macro to test that things work
like expected.

Regards, and thank you for your reply,
Ralf