On Fri, 2010-08-27 at 00:07 +0200, Enrico Weigelt wrote:

> > > They still use the conceptionally broken AC_TRY_RUN(). [1]
> > > So, the chain is broken, everything beyond that cannot be built.
> > > Period.
> > 
> > That is really a problem in your build environment then. 
> 
> No, it's a general misdesign. In 99.99% totally unncessary.

If there are unnecessary calls to AC_TRY_RUN in the glib configure
script, then I suggest that you file bug reports for them and attach
patches. I am pretty sure that the glib maintainers will happily accept
such patches if the calls to AC_TRY_RUN can really be avoided.

> > You can easily feed the expected result of AC_TRY_RUN tests 
> > to the configure script so that it will not attempt to run 
> > the tests.
> 
> Easily ?!
> 
> By analyzing the configure.in and all related m4 files, watch out
> for the AC_TRY_RUN calls, find out what they really should do,
> tweak them to produce the appropriate result - for each specific
> target - w/o trying to run anything. For each single package in
> all required versions. Basicly this means maintaining full 
> downstream-branches, one per package+version+target.
> That totally undermines the whole purpose of autoconf.
> 
> > Pretty much any decent build environment out there has 
> > support to build glib2.
> 
> Which of those are doing clean crosscompiling ?
> How much work to the package maintainers have to spend into 
> each single upstream release, per individual target configuration ?

As far as I can see the effort spent in buildroot to maintain glib2 is
marginal. I have done several glib2 updates myself and it pretty much
boiled down to increasing the version number. It is not that glib would
introduce new AC_TRY_RUN calls every now and then. There are a handful
of such calls and as far as I can see they are all needed.

The good thing is that once you have glib cross-compiled, you are
working in an environment that allows you to write portable code without
the effort of using #ifdefs and feature checks for all platforms.


Sven



Sven


_______________________________________________
directfb-dev mailing list
directfb-dev@directfb.org
http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-dev

Reply via email to