Re: [RFA/python:1/2] Add support for --with-pythondir.

[Doug Evans <dje@google.com>]

On Thu, May 27, 2010 at 1:13 PM, Tom Tromey <tromey@redhat.com> wrote:
>>>>>> "Joel" == Joel Brobecker <brobecker@adacore.com> writes:
>
> Joel> 2010-05-26  Tom Tromey  <tromey@adacore.com>
> Joel>             Joel Brobecker  <brobecker@adacore.com>
> Joel>         * configure.ac: Add handling of --with-pythondir.
> Joel>         * configure, config.in: Regenerate.
> Joel>         * python/python.c (_initialize_python): Add variable "pythondir"
> Joel>         inside module gdb.  Initialize variable gdb.__path__ and execute
> Joel>         __init__.py if found in subdirectory "gdb" of the pythondir.
>
> FWIW I read it and it seems ok to me.
> Thanks for doing this.

fwiw, I still have some concerns.

Support doesn't have to be included in the patch, but IWBN to have a
plan for how to support .so's associated with gdb python modules in a
way that follows Python's scheme: i.e. .py*'s go in
$prefix/lib/pythonX.Y and .so's go in
$exec_prefix/lib/pythonX.Y/lib-dynload [i.e. architecture independent
files go in $prefix and architecture dependent files go in
$exec_prefix].
If the plan is to add --with-python-sodir [or some such] if/when it's
needed, great.  Or if the plan is to decide to just not support it,
that's probably ok too.  I'd just like to put in the thought now.

Also, do we want to name this --with-gdb-pythondir? [and do we want to
rename --with-python to --with-gdb-python?]
The idea being that if gdb was ever configured with something else
that had its own with for --with-python*, then gdb's wouldn't collide.
I don't know if it's that important, just thought I'd bring it up.
Apologies if it's already been discussed and decided and I've
forgotten.