oops, forgot to cc the list.
attached mail follows:
On Wed, Jul 16, 2003 at 12:32:19AM -0700, Greg Stein wrote:
> I very much disagree. We tried that with the Python bindings and using
> distutil. We got screwed. Arguably, the Perl MakeMake thingy is better, but
> I look at your sample Makefile.PL and still see a lot of problems. There are
> a log of things that need to be passed into a proper build of SWIG
> extensions. All kinds of paths, library names, version numbers, etc.
>
> No... we should get the sucker integrated into the standard build system.
> I'm torn on whether to include the Makefile.PL even temporarily. As long as
> it is present, nobody will have any impetus to fix the real build system.
>
> I'm guessing that you'll get a long ways just by doing some simple copy and
> paste in the top-level Makefile.in for the swig-related symbols. Then adjust
> the swig-languages option into build.conf and see what happens :-)
I'd rather leave these for people more familiar with the build system and
libtool. (i don't really like libtool as I tried to fix some stuff with it
before) Of course, we'll have to wait for someone also interested in the perl
bindings and has incentive :)
people should be able to watch what the compiling and linking commands MakeMaker
generates and fix that with build.conf. one thing worth noted would be that
we should tell libtool to generate .${DLEXT} files, where DLEXT could be get from
perl's Config module.
as for now I'll probably use the Makefile.PL for myself and people who wants to
test now, as I feel more comfortable testing stuff with the blib thingy MakeMaker
does. and I could spend more time finishing the typemaps.
Cheers,
CLK
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Wed Jul 16 11:55:16 2003