Re: Allow py-compile to ignore DESTDIR

[Alexandre Duret-Lutz]

>>> "Toshio" == Toshio  <address@hidden> writes:

 Toshio> Hi,
 Toshio> I've been building RPMS of my python program and found that py-compile
 Toshio> is encoding the value of DESTDIR into the byte-compiled .pyc and .pyo
 Toshio> files.  

Good catch!

 Toshio> I put together this small patch to py-compile and
 Toshio> python.am to solve this.  

The change is large enough to require a copyright assignment to
the FSF before we can include it.  Would you be willing to do
that?  (I'll mail you the form separately.)

The patch also lacks a ChangeLog entry, and if possible a test
case (proving it works, and ensuring nobody breaks this later).

 Toshio> It creates a --destdir [DIR] option to py-compile that
 Toshio> complements --basedir.  Using --basedir specifies an
 Toshio> additional path that is encoded on the beginning of the
 Toshio> filenames whereas --destdir specifies an additional
 Toshio> path to tack on the front of the filenames only for the
 Toshio> purpose of finding the file to compile.

Why not.  My preference would be to use --destdir to specify $(DESTDIR)
as in

   $(py_compile) --destdir "$(DESTDIR)" --basedir "$(%NDIR%dir)" $$dlist   

to avoid the repetition of directories and make the use of
--destdir clearer.  But that's not a big deal.

Also it seems an easy way to simplify the python part of the
script is to do

  test -n "$destdir" && cd "$destdir"  # assuming the above semantics
  test -n "$basedir" && cd "$basedir"

before actually running python.  Then the python part does not
need to know about $destdir, and probably only the py_compile
line needs to be changed to

  py_compile.compile(file, file + 'c', path)

Does this sound correct?
-- 
Alexandre Duret-Lutz