Re: Bug#622251: pu: package python-apt/0.7.100.1+squeeze1
On Mo, 2011-04-11 at 19:20 +0100, Adam D. Barratt wrote:
> On Mon, 2011-04-11 at 12:46 +0200, Julian Andres Klode wrote:
> > We'd like to upload python-apt 0.7.100.1+squeeze1 (or without
> > +, whatever is preferred) to stable. It contains the changes
> > the ftpmasters seem to need for multi-arch support and xz
> > packages in the archive.
>
> I'm sure this came up in earlier discussions regarding the multi-arch
> change, but I can't find it right now. So far as I can see, this:
>
> + * python/apt_pkgmodule.cc:
> + - strip multiarch by default in RealParseDepends
> + - add optional parameter to allow parse_depends() to keep the
> + multiarch parameter
>
> is an API change - i.e. the version of python-apt currently shipped in
> Squeeze does /not/ strip multiarch specifiers by default? Do we know if
> any of the reverse dependencies in Squeeze use ParseDepends() and might
> be affected by the change?
Since we do not support multi-arch in python-apt in squeeze (python-apt
raises exceptions on multi-arch systems in squeeze), there should not be
any reverse dependencies using it. We're shipping it in unstable since a
few weeks and no one complained yet. And it's going to be shipped in
Ubuntu natty.
> As a side note, if the default in apt's debListParser::ParseDepends() is
> apparently wrong, is there a reason that's being worked around in
> python-apt, rather than fixed in apt directly?
It's fixed in apt 0.8.13.2. Yes, I am not happy with changing the
default value for a function argument (especially without changing
soname, as that's formally an ABI break). It's been done by David
Kalnischkies, I'm sure he has a reason for it.
--
Julian Andres Klode - Debian Developer, Ubuntu Member
See http://wiki.debian.org/JulianAndresKlode and http://jak-linux.org/.
Reply to: