Re: Python 3 support in Squeeze
On Fri, 2010-09-03 at 20:50 +0200, Piotr Ożarowski wrote:
> [Adam D. Barratt, 2010-09-03]
> > On Sat, 2010-08-21 at 00:44 +0200, Piotr Ożarowski wrote:
> > > We plan to upload such NEW versions of Squeeze packages to experimental²
> > > (one of ftp-team members, who is a member of our team, will help us with
> > > that), test it there and if everything will be ok - upload to unstable
> > > and ask for freeze exception³.
> >
> > Has the initial side of that process begun yet (i.e. the preparation of
> > the new packages and uploads to experimental)?
>
> I can have my packages ready really soon (I did some initial work). I
> can also prepare patches for the ones I listed previously (I took a look
> at them before sending my mail). Then I plan to write some instructions
> so that more people will join.
I have to admit that I was hoping / expecting that you'd have begun
staging at least some of the packages in experimental already, to have
given time to test them and be able to migrate them to unstable quickly
if / when we approved that.
> > > Please note that in most (all?) cases 2to3 tool (which converts
> > > python2.X code to python3.X one) will have to be used (again, no new
> > > upstream versions) so patching the code in Squeeze (security bugs, etc.)
> > > will not have to be done twice (at least in most cases).
[...]
> > How confident are you in the code produced by 2to3?
>
> all packages I listed were checked by upstreams (and they claim 2to3 did
> a good job). I don't think it's a good idea to convert sources that
> upstream didn't test with Python 3 and I don't plan to convert such
> packages.
So upstream have verified the result of 2to3 on their software?
> > If one of the packages involved
> > needed a security update in stable, what are the chances that 2to3 would
> > fail to convert the patched code, produce a script which doesn't run,
> > doesn't work or (hopefully a very unlikely worst case) doesn't properly
> > fix the bug?
>
> how about sending such patches to debian-python@l.d.o? You can also ping
> me directly.
On a related note, will the build system allow for the result of 2to3 to
be easily overridden? For example, if whilst building an update the
security team discover that there's a problem with the 2to3-generated
patch, will they be able to add a fixed patch to the package and have
that used in preference?
Regards,
Adam
Reply to: