Re: gnustep make packaging issue

Andreas Fink via Discussion list for the GNUstep programming
environment wrote:
> On 02.10.2025 12:34, R Frith-Macdonald wrote:
> > That is, your output suggests that gnustep-make is doing the right
> > thing, and certainly defining  GNUSTEP_RUNTIME=1 is correct for
> > using the new runtime (ie NOT the GNU one that comes with GCC).
> 
> well base/gui/back/corebase now compiles and install
> 
> now my own configure.ac  makes strange things. it fails at the check
> for cross compilation

I guess debugging this requires the relevant configure.ac snippet and
(at least some part) of the corresponding config.log.

> reproducing that test by hand fails that it fails after including
> Foundation/Foundation.h that it doesnt know about things like NSArray
> etc

Just speculating: this may be due to improper usage of #include in
AC_COMPILE_IFELSE and friends.  You should probably call AC_PROG_OBJC
and then wrap all subsequent Objective-C specific checks within
AC_LANG_PUSH/AC_LANG_POP.  If you're using obsolete macros, please
upgrade, they're obsolete (read: broken) for a reason and won't be
fixed, ever.

> interestingly if I call ./configure with --host
> aarch64-unknown-linux-gnu which is the architecture I'm running
> on. it deso not fail which i find puzzling.

Not puzzling at all.  It might indicate a bug in your build system
(just speculating, I don't have sufficient information to maintain
such claim).  Note that gnustep-make has very different notion of
build/host/target compared to GNU Autotools (perhaps that's exactly
what's biting you?).

> But whats not puzzling with autotools :)

This is the usual complaint from people who didn't bother to read the
fine manuals.  I don't want to sound abusive, so I apologize if you
felt it that way (it definitely was not intended).