[Luke W Faraone, 2017-01-30]
> In bug #744741[1], we have a report where a lack of ``egg-info`` metadata
> breaks
> both ``pip``'s installation detection and packages that use ``pkg_resources``
> to
> discover dependencies.
>
> Usually, the answer would be simple: ship the ``egg-info`` metadata as part of
> the package. But PySide is distributed upstream[2] and via PyPI[3] as one
> monolithic package. Debian splits that out into 14 packages, ``python-
> pyside.phonon``, ``python-pyside.qtcore``, ``python-pyside.qtdeclarative``,
> etc.
>
> So, should we ship the egg-info files in a common package, as Barry
> suggested[4], and make each of the submodules depend on it? This has the
there's already python-pyside and python3-pyside that depend on all
other subpackages. Upstream expects pkg_resources.require('PySide') to
confirm that all of them are installed (as otherwise there would be
more egg-infos) so the answer is simple: ship it in python-pyside and
python3-pyside.
> unfortunate side-effect of breaking third-party packages that attempt to
> detect
> whether PySide is installed.
>
> Alternatively, we could only distribute it as part of ``python-pyside`` (a
> metapackage), but this would require patching to some Debian-distributed
> packages such as ``yubikey-piv-manager``.
yubikey-piv-manager can either depend of python-pyside or remove PySide
from setup.py in order to keep (in Depends) python-pyside.qtgui and
python-pyside.qtnetwork only, yes
> A third option would be to ask upstream to split out the packages as we have
> done -- that would resolve the conflict in this instance, but not the general
> issue, and would probably take a lot of effort (or be rebuffed).
even more eggs? Please no.
--
Piotr Ożarowski Debian GNU/Linux Developer
www.ozarowski.pl www.griffith.cc www.debian.org
GPG Fingerprint: 1D2F A898 58DA AF62 1786 2DF7 AEF6 F1A2 A745 7645
