Daira Hopwood <da...@leastauthority.com> writes: > The "standard interface" for 'python setup.py install' does not do what > I or you or anyone else wants it to do.
What I meant is
pkgsrc has infrastructure for dealing with packages that use
setuptools/distutils. That's all debugged. I imagine all the other
packaging systems have the same situation. So the net result of 'make
package-install' (or whatever the local incanation is) is right, even
if some of the steps are goofy. But the steps are goofy in the same
way, so it all works. Changing one package that uses setup.py to do
something different than what the others do is a regression.
What I expect 'python setup.py install' to do is put the tahoe bits in
site-packages, and not to do anything about dependencies. The
dependencies were already installed because the package declared a
dependency on the package that has each dependency.
The actual call to setup.py install from 'make install' of the
tahoe-lafs package looks like:
[lots of environment variables]
/usr/pkg/bin/python2.7 setup.py "install" -c -O1 \
--single-version-externally-managed \
--root=/usr/pkgsrc/filesystems/tahoe-lafs/work/.destdir
because it's not actually installing into the real system, but into a
destdir which will then be tarred up to make the package.
(I realize --single-version-externally-managed is critical. Debian has
the same approach as pkgsrc: http://wiki.debian.org/Python/FAQ )
pgp5Et5HhLXjq.pgp
Description: PGP signature
_______________________________________________ tahoe-dev mailing list tahoe-dev@tahoe-lafs.org https://tahoe-lafs.org/cgi-bin/mailman/listinfo/tahoe-dev
