[PATCHES] Build pytalloc for two Python versions at once, port to py3

[Jelmer Vernooij]

On Thu, Mar 19, 2015 at 04:07:24PM +0100, Petr Viktorin wrote:
> On 03/19/2015 02:26 PM, Jelmer Vernooij wrote:
> >On 19 March 2015 9:10:38 am GMT+00:00, Petr Viktorin <pviktori at redhat.com> wrote:
> >>
> >>Are you now saying that supporting Python 3 in thestand-alone libraries
> >>
> >>is a bad idea?
> >I don't think that is a bad idea for the standalone libraries. We're concerned
> >about the longevity of dual-python support though, especially in Samba itself.
> >
> >>Can you tell me what hoops are a burden, so I can try removing them?
> >Supporting two python versions at the same time and all the issues that
> >come with that. I think the specifics have been well covered in this thread.
> 
> Somehow, the discussions always turns to potential Python3 support in Samba
> itself. Unless I misunderstood, most of the specifics mentioned in the
> thread are about that.

> Not that it's not a good discussion -- I'm not saying I don't want Samba
> ported eventually, and I'm starting to put into action some ideas to make
> the ecosystem better for Samba when it does decide to port -- but the main
> thing I'm asking for in this thread is patch review. It's hard to separate
> comments on eventual porting of Samba from comments on the porting
> stand-alone libraries. If it looks like I'm stubbornly ignoring some issues
> raised on the patches, this might be what I misunderstood.
Fair enough; let's focus on the short-term goals (dual-python support in
the standalone libraries) in this thread. I still think that is a worthwile
goal, independent of what ends up happening for Samba itself, but we
can have that discussion later.

> Is there anything I can do to make the build system and talloc patches
> better?
> Should I first complete porting the other stand-alone libraries, and present
> a giant set of patches then? In other words, is having talloc ported useless
> without the other stand-alone libraries being ported as well? I know the
> other libraries will have issues with unicode/bytes; pytalloc does not.
> I think a giant patchset would complicate things for reviewers as well as
> for me, but if you'd prefer it, I can go hack on it. Even then, if you have
> any comments on the changes at hand, I'd like to hear them so I don't have
> to rebase so much down the road.
I agree. I think a patchset just for talloc is fine.

> If the extra header is too much, I can move the macros into the c files.
> (I'd personally prefer not to do that, so that it's easier to re-use them
> consistently across all standalone libraries. But I can move them if you
> wish.)
My strong preference is for inlining the macros. If somebody else (Andrew?)
wants to review this patchset with the header I won't stand in the way,
since there seems to be disagreement about this point, but I won't review.

I need to take a closer look at the waf changes, but other than that and
the py3compat.h file this patchset looks good to me.

Cheers,

Jelmer