Laurent,

On Thu, Dec 2, 2010 at 5:58 PM, Laurent Gautier <lgaut...@gmail.com> wrote:
> Thanks for the patch, however at first sight I am surprised that is it
> really working for you.
> Wouldn't
> '^(?P<extra_link_args>-Wl.+)$'
> be better ?

yes, you're probably right... to be honest, I hadn't digged through
your code much.
I just  realized that setup.py bails out because it doesn't recognize
a -Wl,<...>
option, and so I let it pass. I also hadn't encountered named patterns
before (and
was reasonably happy when I got it working). Is it right that not including the
extra_link_name results in the option not being used eventually? I
guess it works
in my case because I have specified the MKL libraries at least twice when
compiling R (once with --with-blas, and a second time with --with-lapack)
when  running the configure script, so the grouping probably isn't
even necessary.

> Yes: this is making sure that nothing unexpected is happening, and something
> does abort. This procedure helped solve installation-specific issues quicker
> than it would have otherwise (especially when the problems are reported by
> users with limited experience/interest in the details of compiling Python
> extensions). It is also making the code for parsing of paths and libraries
> somewhat clear and explicit, and let setup.py rely rather simply on the
> package building infrastructure in distutils. Also the people with
> less-common compiling options are hoped to able to identify precisely a
> problem they might encounter and provide a patch.

That makes perfect sense.

> The detail is that distutils wants for example an include _directory_ rather
> than '-I/my/include/are/there', requiring us to parse the directory out of
> the string.

Parsing of arguments is fine. Still, one could argue that additional options
not related to directories or library names shouldn't stop the build immediately
just because they haven't been foreseen...

> Having that said, I might have missed something and I would read with great
> interest a patch to reform the setup.py.

Oh - your kind answer to my question already shows that you have thought
much deeper about distutils than I ever could. I'm usually just glad when I can
figure out (for the n-th time) how to specify whatever option to the compiler I
happen to pass on.

Have you ever considered using distribute (which stated off from setuptools and
seems to be the future distribution standard)? Its maintainer, Tarek
Ziade, who also
works on distutils2, recently voiced an interest in improving the
support for compiling
(see 
http://tarekziade.wordpress.com/2010/11/07/distutils-2-alpha-4-work-in-progress/),
and asked for input from scientific python communities. Maybe your insights and
experiences could help?

In any case, thanks for your excellent work on Rpy!

   Christian.

------------------------------------------------------------------------------
Increase Visibility of Your 3D Game App & Earn a Chance To Win $500!
Tap into the largest installed PC base & get more eyes on your game by
optimizing for Intel(R) Graphics Technology. Get started today with the
Intel(R) Software Partner Program. Five $500 cash prizes are up for grabs.
http://p.sf.net/sfu/intelisp-dev2dev
_______________________________________________
rpy-list mailing list
rpy-list@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/rpy-list

Reply via email to