Re: [gentoo-python] Reducing exceptionality cases for PYTHON_SINGLE_TARGET

[Michał Górny]

Dnia 2014-10-29, o godz. 14:17:29
Ian Stakenvicius <axs@g.o> napisał(a):

> > PYTHON_TARGETS="python2_7 python3_3 python3_4" \
> > PYTHON_SINGLE_TARGET="python3_3" \ emerge -av sys-devel/distcc
> >
> > These are the packages that would be merged, in order:
> >
> > Calculating dependencies |
> >
> > !!! Problem resolving dependencies for sys-devel/distcc ... done!
> >
> > !!! The ebuild selected to satisfy "distcc" has unmet
> > requirements. - sys-devel/distcc-3.1-r9::gentoo USE="gtk -avahi
> > -hardened -ipv6 -selinux -xinetd" ABI_X86="64"
> > PYTHON_SINGLE_TARGET="-python2_7" PYTHON_TARGETS="python2_7"
> >
> > The following REQUIRED_USE flag constraints are unsatisfied:
> > python_single_target_python2_7
> >
> > The above constraints are a subset of the following complete
> > expression: python_single_target_python2_7? (
> > python_targets_python2_7 ) exactly-one-of (
> > python_single_target_python2_7 )
>
>
> There are two ways to deal with this in-ebuild -- the first would be
> to adjust the ebuilds to python-r1 but not multibuild them, which
> means that the ebuilds are python-single but not actually implemented
> in a python-single way.

This is an easy way to get package broken once someone adds python3
support.

>  The other is to adjust python-single-r1 so
> that if there is only one possible implementation (ie, PYTHON_COMPAT
> just has one entry, or only one such entry is in _PYTHON_ALL_IMPLS)
> then the PYTHON_SINGLE_TARGET setting is ignored.
>
> The main issue that this may cause is that the python_single_target_*
> use dependencies are no longer fully deterministic at all times, and
> are rather now determined based on the supported python implementation
> list in _python_impl_supported() from python-utils-r1 -- ie, when that
> list changes, some packages may suddenly require PYTHON_SINGLE_TARGET
> be set while others just-as-suddenly won't need it.  This may mean
> some extra rebuilds for users on "emerge -N", especially as the flags
> are dropped.

This is ugly.

I was rather thinking of making all eclasses put +impl in IUSE when
only one implementation is supported.

-- 
Best regards,
Michał Górny