Hi Matthias,
I am subscribed to Debian automake1.10 bugs, no need to Cc: me.
(Incidentally, does it work if I reply to submit@ instead of the bug
number?) (If one can subscribe to ubuntu packages' bugs somewhere, I'd
appreciate a pointer to that, too.)
* Matthias Klose wrote on Wed, Apr 15, 2009 at 12:17:57PM CEST:
> The current implementation of the AM_PATH_PYTHON passes the unexpanded
> '${prefix} and ${exec_prefix} to get_python_lib(). The documentation of
> get_python_lib() says:
>
> If 'prefix' is supplied, use it instead of sys.prefix or
> sys.exec_prefix -- i.e., ignore 'plat_specific'.
>
> IMO the m4 macro assumes that get_python_lib() just passes this and doesn't
> use
> the prefix to do anything, which is at least wrong for Debian's python2.6 and
> python3.1 in experimental, where we determine the name of the site directory
> depending on the location ('dist-packages' for a prefix of '/usr' or
> '/usr/local', 'site-packages' for anything else, like custom locations or
> installations in python-virtualenv). Just passing the unexpanded value,
> get_python_lib() cannot decide which name to use for the site directory.
>
> The idea of this patch is to pass the real prefix, and then replace it again
> with the unexpanded value, so that the unexpanded value ends up in the
> automake
> variables according to the documentation. Afaics this patch doesn't have an
> effect on any standard python installation, but correctly determines the
> location of the site directory on Debian.
Thanks again for the report. I am willing to integrate this in upstream
Automake, but I need some way to "see the bug happening" (I have a
Debian system to try out on), to fully understand it and to ensure that
the bug this fixes remains fixed. So can you please describe a short
setup that I can try to mangle into an addition for Automake's
testsuite?
(For nontrivial patches, integration into upstream will require
copyright papers, as I guess you well know, so if you are not ready
to deal with that, it's better if you describe such a test to me rather
than produce a patch for a testsuite addition; details off-list if you
like.)
Thanks,
Ralf
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
