On Wed, Aug 29, 2012 at 6:14 PM, Michał Górny wrote:
> On Wed, 29 Aug 2012 18:05:19 -0400 Mike Frysinger wrote:
>> On Wed, Aug 29, 2012 at 6:02 PM, Michał Górny wrote:
>> > On Wed, 29 Aug 2012 17:50:16 -0400 Mike Frysinger wrote:
>> >> On Wed, Aug 29, 2012 at 5:42 PM, Michał Górny wrote:
>> >> > In other words, pkg-config is only used when no other criteria
>> >> > allows it to classify the particular .la file as suitable for
>> >> > removal or not. Sadly, it's rather, ehm, unfriendly to ebuild
>> >> > developers who obviously don't even read the relevant part.
>> >> >
>> >> > Do you have any ideas how we can improve that?
>> >>
>> >> before the func executes pkg-config, run `has virtual/pkgconfig
>> >> ${DEPEND}` and spit an eqawarn if it's not found
>> >
>> > Ciaran will shot at me for doing that.
>>
>> it isn't violating anything and can find real bugs. i don't see a
>> problem here.
>
> It is violating the Holy PMS.
does it actually ? are DEPEND variables not allowed to be expanded in
pkg_* src_* funcs ?
we could probably add a similar check to autotools.eclass: grep for
PKG_PROG_PKG_CONFIG and check ${DEPEND}
>> >> > One thing that comes into my mind is finally making pkgconfig
>> >> > a required, implicit part of toolchain (or @system). Since we
>> >> > have pkgconf now, this is more feasible than before.
>> >>
>> >> i don't think making it part of the toolchain makes sense. i'd
>> >> rather not add it to @system simply to keep a few packages from
>> >> sometimes failing.
>> >
>> > I'd add it to @system because a lot of packages actually need to
>> > DEPEND on pkgconfig because they use libraries using .pc files. And
>> > the number is going to increase, hopefully.
>>
>> sure, but keeping things in @system doesn't make much sense:
>> - there's a penalty (as noted in old threads)
>> - it isn't actually required at runtime, so it's bloat on reduced
>> systems
>
> I think it's practically the same as compiler.
that isn't a bad view point, but for the purposes of this discussion,
i don't think it's relevant :)
> Also, a quick look at !ddep shows over 7000 reverse dependencies. That
> looks like a bigger penalty to me.
if we had a @build-system, you might be able to convince me. but we
don't. so the number of packages here doesn't matter as it's an
invalid implicit RDEPEND.
-mike
