Re: Problem with AC_CHECK_FUNCS when cross compiling

[Thies Jochimsen]

Dear Ralf,

sorry, it was my mistake. It was a complex interplay between the cross 
compiler (for VxWorks), a (lousy) wrapper script for the compiler, and the 
new way autoconf checks for the executable bit off binaries. I fixed the 
wrapper and now everything is fine...

Thanks, however, for your help, Thies


On Dienstag, 28. April 2009, Ralf Wildenhues wrote:
> Hello Thies,
>
> * Thies Jochimsen wrote on Mon, Apr 27, 2009 at 02:03:20PM CEST:
> > I have just upgraded to autoconf 2.63 (from 2.61) and I am now struggling
> > with way the AC_CHECK_FUNCS behaves when cross compiling:
> > In the past, I have used AC_CHECK_FUNCS to test whether certain functions
> > are available on the the target host. But with 2.63, the test in
> > AC_CHECK_FUNCS now, additionally, returns true if $cross_compiling is set
> > to "yes" (which is always true in my case) and, thus, always declares a
> > HAVE_SOMEFUNCTION, regardless whether the function is actually available.
> >
> > I understand that there was a reason for adding the extra check for
> > $cross_compiling to AC_LINK_IFELSE:
> > http://lists.gnu.org/archive/html/autoconf/2007-03/msg00085.html
> >
> > Is there a convenient way to check for functions available on the target
> > host, even when cross compiling?
>
> I'm not sure I understand; I cannot connect your description in the
> first paragraph with the mailing list reference below that.
> Can you be more specific and show us an full example that produces
> the wrong answer for you, or different answers between Autoconf 2.61 and
> 2.63, including all commands you need to get there?
>
> A small example would be something like this; maybe you can modify this
> so that it exposes the issue:
>
> cat >configure.ac <<'END'
> AC_INIT
> AC_CONFIG_HEADERS([config.h])
> AC_PROG_CC
> AC_CHECK_FUNCS([doesnotexist])
> AC_OUTPUT
> END
>
> autoconf
> autoheader
> ./configure
> grep HAVE config.h
>
>
> Thanks,
> Ralf