On Wed, 22 Jul 2009 16:52:55 +0900, David Cournapeau <da...@ar.media.kyoto-u.ac.jp> wrote:
[snip] Installing libraries which are not python extensions is indeed quite difficult ATM, I faced the same problem in numpy recently. Since in numpy, we have our own distutils extensions, I hacked something to make it work. A lot of code is related to our needs to be able to reuse to libraries (with link options and all), and otherwise: - I patched the Distribution class to add one member which keeps track of the installed libraries (dll in your case) - I added an install_clib command: this takes care of adding the files to the record file, etc... - I added the in-place option for build_clib command: this is necessary to make sure develop command works Feel free to reuse this if you want (and can! It is butt-ugly, but not worse than average distutils code :) ). http://github.com/cournape/numpy/tree/install_clib
Cool, thanks! I'll take a look at this. Jean-Paul _______________________________________________ Distutils-SIG maillist - Distutils-SIG@python.org http://mail.python.org/mailman/listinfo/distutils-sig
