[Bug 526126] Review Request: python3 - Python 3.x (backwards incompatible version)

bugzilla at redhat.com bugzilla at redhat.com
Tue Oct 13 21:26:24 UTC 2009


Please do not reply directly to this email. All additional
comments should be made in the comments box of this bug.


https://bugzilla.redhat.com/show_bug.cgi?id=526126





--- Comment #8 from Andrew McNabb <amcnabb at mcnabbs.org>  2009-10-13 17:26:22 EDT ---
Dave, thank you very much for your insightful thoughts.  I agree with you on
almost all points.  The important thing for me is getting a working RPM that
can be installed alongside Python 2.  As you noticed, a lot of things in my
specfile exist purely to be similar to Python 2 (although I changed anything
that seemed wrong or inapplicable).  I wouldn't have any hard feelings over
which specfile is used.  However, there are some areas where I think mine is
more "right" and other areas where I think ivazquez's is probably better. 
Obviously, we should want to combine the best of both.  Here are a few specific
thoughts:

1) I think that the name python3 is clearly preferable to python3000.

2) I'm confused by the .list thing.  The .list files are really short, and I
don't see how they make it any more clear.  In fact, I think they make the
specfile more difficult to read.  Is this a standard way to do things?

3) The patch python-3.1.1-config.patch follows the approach of Python 2.  This
is really important (whoever did Python 2 really knew what they were doing). 
Python's default build configuration (without the patch applied) will fail
silently if a module can't be built.  This is really evil.  It means that if a
dependency is missing, then files will just disappear.  This patch gets rid of
silent failures by explicitly specifying which modules must be built.

4) The patch python-3.1.1-lib64.patch ensures that Python is installed into
/usr/lib64 (as per Fedora's packaging guidelines).  It looks like the
python3000 specfile does the same work with sed instead of a patch.  I don't
know which approach is better, and I could be easily persuaded either way.  I
picked to make a patch to more closely match the Python 2 specfile, but I can
definitely see some advantages to using sed.

5) I have no strong feelings on the package description or build dependencies. 
However, I don't get the disparaging remarks about Tkinter.  Not that I have
any attachment to Tkinter, but I'm not aware of it being considered in any way
deprecated.

6) The configure option --with-wide-unicode seems to be the correct spelling
for the option, where --enable-unicode=ucs4 is the old way to say it.

7) By the way, the python3000 package defines a python3000-libs subpackage, but
the python3 package just includes the .so file in the main package.  When I
tried to have a python3-libs package, it ended up being a prerequisite for
python3 anyway, so I couldn't see the point in keeping it separate.  However,
it's entirely possible that I did something wrong to make this happen. :)

ivazquez, do you have any thoughts on these issues?  Is there any particular
reason that you didn't submit your package for review?  Thanks for your work!

-- 
Configure bugmail: https://bugzilla.redhat.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.




More information about the Fedora-package-review mailing list