I've found pyparsing rather brittle between revisions in the past, hence
the desire to have a local copy. I don't know if recent versions have
stabilized -- given that we have something that works, I'm not too keen
on tinkering with it.
Not knowing much about packaging myself, I think Debian should only go
forward with using an external dependency if an *exact* version of
pyparsing can be specified, rather than >=. Or at the very least, if a
different version of pyparsing is applied, one needs to make sure that
the mathtext examples all pass unchanged.
Mike
Sandro Tosi wrote:
> Hi guys,
>
> On Fri, May 29, 2009 at 11:51, Daniel Watkins
> <da...@da...> wrote:
>
>> Package: python-matplotlib
>> Version: 0.98.3-5
>> Severity: normal
>>
>> python-matplotlib installs its own copy of pyparsing.py when it should
>> in fact be using the copy that is shipped in python-pyparsing.
>>
>
> We've just receive this bug report about the internal copy of
> pyparsing included in mpl.
>
> The situation in Debian is:
>
> Stable 1.5.0-1
> Testing 1.5.1-2
> Unstable 1.5.2-1
>
> Currently mpl ship:
>
> $ grep "^__version" lib/matplotlib/pyparsing.py
> __version__ = "1.5.0"
> __versionTime__ = "28 May 2008 10:05"
>
> In the changelog I can see:
>
> $ egrep -A2 "2007-11-09.*pyparsing" CHANGELOG
> 2007年11月09日 Moved pyparsing back into matplotlib namespace. Don't use
> system pyparsing, API is too variable from one release
> to the next - DSD
>
> So there seems to be a reason for this "private" copy. The question
> is: is this reason still valid nowdays? should we (at least packagers)
> remove the private copy and rely on the system pyparsing (or at least
> introduce a "check if system has pyparsing, if not fallback on
> private" wrap)?
>
> I haven't checked, but maybe you already know the answer :)
>
> Cheers,
>
--
Michael Droettboom
Science Software Branch
Operations and Engineering Division
Space Telescope Science Institute
Operated by AURA for NASA