Re: [Boost-testing] Python + Solaris + -lrt

From: David Abrahams (dave_at_[hidden])
Date: 2007-03-25 12:06:23

on Sun Mar 25 2007, "Caleb Epstein" <caleb.epstein-AT-gmail.com> wrote:

> See:
>
>     http://tinyurl.com/ynjs5o
>
> This test makes use of facilities in Python that require linking with
> -lrt on Solaris.  I think we have covered this ground in the past,
> though I am having difficulty finding a precise reference.  This
> thread is a similar one where the Jamfiles were requesting a link with
> non-existent-on-Solaris libutil (a BSD-ism I believe):
>
>    http://thread.gmane.org/gmane.comp.lib.boost.testing/1010/

The code in python.jam *should* be causing -lrt to be added.  Let me
look at this again...

> Would it not make more sense to ask Python itself what libraries it
> needs, rather than having this encoded in the Jamfiles?  It looks to
> me like this might work:
>
>     #!/usr/bin/env python
>     from distutils import sysconfig
>     print sysconfig.get_config_vars()['LIBS']
>
> This outputs:
>
>      -lresolv -lsocket -lnsl -lrt -ldl

Whoa, I've been looking for something like this, but didn't know where
to find it.

Yes, it would make more sense, for the great majority of cases where a
Python interpreter is available on the system.  I don't know if
there's time for that change right now, though.

> on my platform (SunOS moma 5.10 Generic_Patch_118844-30 i86pc i386 i86pc)
>
> PS.  Looking at the v2 python.jam, I see that the rt library is added
> via an "extra-libs-conditional" directive.  This doesn't appear in the
> link line for this test however.

I had been waiting for these results to update, which is why I wasn't
alarmed about this before.  Eventually I just checked in a trivial
change to boost/python/detail/config.hpp so that incremental testers
would be forced to update.  Now I'm concerned. :(

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com