Re: configure promise + new seg faults

[Alexander Malmberg]

Stefan Böhringer wrote:
> On Tue, 2003-08-12 at 05:21, Adam Fedor wrote:
> > On Monday, August 11, 2003, at 03:53 AM, Stefan Böhringer wrote:
> [..]
> > > I figured that GSConfig.h was empty - and that I now have to run
> > > configure in each of the core components. The promise of having to
> > > configure only once in make seems to have gone. Such changes should be
> > > clearly announced on the list IMHO.
> > >
> > I'm not sure what you mean. You always have had to run configure in
> > each of the core libraries.
> [..]
> Hm,... not so long ago at least it worked to just run configure in make.

IIRC, there have been configure:s in all components for a long time, but
you only really had to run the one in -make. The others would run
automatically. I haven't tested, but this might have been broken.

> Also the compile-all script did not contain configure calls other than
> in make not a long time ago.

According to cvs, the configure:s have been there since the file was
added.

> I'm not aware of any official statement,
> but as far as I understood GNUstep was ment to be rather configure-free
> in that the important variations are checked for when installing make
> and other packages are shielded from configure through a global
> GSConfig.h (or the libraries themselves if end user applications are
> concerned).

I don't think that checking for everything in -make is a good idea, and
given that GNUstep has had configure:s for each component for a long
time, I don't think that's the intention, either. However, automatically
running configure for components other than -make seems like a good
idea, so if that's been broken, it should probably be fixed.

- Alexander Malmberg