# not tagging wontfix given it's assigned to both packages but reassigning
reassign 836535 cross-pkg-config
thanks
Hi,
On Sat, Sep 03, 2016 at 09:41:41PM +0200, Helmut Grohne wrote:
> Please restate[1] your reasons for rejecting the proposed solutions in
> this bug. Failure to do so will result in me NMUing your package
> (according to the usual delayed rules). Then, please agree on an
> approach that works for both of you two.
The bug is in cross-pkg-config by not honouring "standard" pkg-config paths.
I was not, am not and will not do multiarch in my packages.
And moving just the pkg-config file to the multiarch path is doing just
the thing "in half", which I consider also wrong.
But as you apparently want, here the IRC parts:
17:21 < _rene_> helmut: yes, object. definitely. ESPECIALLY for the multi-arch
stuff.
17:22 < Chipzz> the program crashes on "process launch", while gdb works fine
for the same binary
17:22 < _rene_> helmut: I mean, LIBDIR=/usr/lib there is a sure sign this was
deliberate, no?
17:22 < _rene_> helmut: most of my libs have that :)
17:23 < _rene_> helmut: (context: librevenge)
[...]
18:17 < helmut> _rene_: well, it makes packages fail to cross build. please
discuss that with tollef then
18:17 < helmut> _rene_: i.e. either move at least librevenge.pc or fix
pkg-config.
18:19 < helmut> _rene_: the point of a software distribution is that packages
cooperate. saying "no" to that cooperation without a good
reason means you are in the wrong boat.
18:19 < _rene_> no, I don't. it's your crusade to to multiarch, no mine.
18:20 < helmut> _rene_: this is not about multiarch (even though multiarch
would be a simple solution), but about being able to build
stuff from source.
18:20 < _rene_> I cooperate in stuff needed, please don't get me into this
matter, please
18:21 < _rene_> helmut: librevenge builds. and obviusly librevenge.pc is found
by pkg-config otherwise the r-deps wuoldn't build
18:21 < _rene_> or did pkg-config change behaviour and broke a unknown number
of packages?
18:21 < _rene_> # pkg-config --cflags librevenge-0.0
18:21 < _rene_> -I/usr/include/librevenge-0.0
18:21 < _rene_> no, it didn't
18:21 < helmut> no, rdeps (e.g libwpd doesn't find librevenge.pc) don't cross
build
18:21 < _rene_> *cross* build, yes-
18:22 < _rene_> since when are we doing cross builds? I mean, will it ever be
officially done?
18:22 < helmut> there is not much choice for 32bit architectures
18:23 < _rene_> huh? we have 32bit archs? and because? size limits? so be it?
18:24 < _rene_> how would cross-build help there (especially for stuff which is
not cross-buildable?)
18:24 < helmut> you can cross build from a 64bit arch to a 32bit arch
18:24 < helmut> objecting to things like "cross buildable" or "reproducible
builds" on the grounds of "no" is plain stupid
18:25 < _rene_> I am not objecting against them
18:25 < _rene_> I am objecting against multiarch
18:25 < _rene_> if the latter breaks the cross-build attempts as that needs
multiarch, well...
18:26 -!- atoz [[email protected]] has joined #debian-devel
18:26 < p2-mate> we could also drop 32bit archs :)
18:27 < helmut> _rene_: well, if you have no good reason, I suggest that you
discuss that with the ctte.
18:27 < _rene_> there's no "you must use multiarch" ctte discussion
18:27 < _rene_> or resolution, even
18:28 -!- taffit_ is now known as taffit
18:28 < pabs> whats the problem with multiarch?
18:28 < _rene_> that one would cause me a mass-upload of my packages to do
multiarch (even for LO where it doesn't make sense except
having the public libs in a "correct" place), but until then...
18:29 < helmut> _rene_: again, this is not about multiarch. this is about
cooperation of the librevenge maintainer with the pkg-config
maintainer to produce something useful
18:29 < _rene_> no, it's about multiarch for the goal for cross-building
18:29 < _rene_> right now it's useful and the pkg-config file *is* found
18:30 < _rene_> in "standard" builds
18:30 < helmut> _rene_: would you be ok with moving just the pkg-config file?
(not the libs)
18:30 < _rene_> no, what for? either do it completely or not.
18:30 < _rene_> and I decided for not.
18:30 < helmut> because pkg-config does not consider /usr/lib/pkgconfig
18:30 < _rene_> it does in standard builds
18:30 < _rene_> maybe not in cross-builds, but..
18:31 < _rene_> .oO ( maybe I should just give in and do the mass-upload when I
have a Saturday completely free and can do the 100ish packages )
18:32 < helmut> I don't think we need to convert everything, just the stuff
that is "important" in some way (e.g. key package, high popcon,
many rdeps). librevenge happens to be important under those
metrics
18:33 -!- nikos [[email protected]] has joined #debian-devel
18:33 < _rene_> and note that dh_make even now does multiarch paths in .install
but no Multi-Arch: something in control
18:34 < helmut> I know. and multiarch paths is what I need here. not multiarch
headers
18:34 < helmut> just changing LIBDIR is sufficient here.
18:34 < _rene_> helmut: I (almost) exclusively mantain a key package and MANY
of it's r-deps (just packaged because of that one,
initially)..., so...
18:35 < _rene_> helmut: where we are again at "either do it completely or
not.", otherwise it's just a half-baked solution.
18:35 -!- piper [[email protected]] has joined #debian-devel
18:35 < helmut> it's a transition, yes.
18:36 -!- atoz [[email protected]] has quit [Quit: atoz]
18:36 < helmut> we have about 10000 source packages that build arch-dep
packages. about 5000 have cross-satisfiable build-depends and
we know that about 1000 are cross buildable already.
18:37 -!- gvaillan [~gvaillan@2001:630:12:2e1e:c57f:2e88:f339:ee78] has joined
#debian-devel
18:38 -!- zwenna [[email protected]]
has joined #debian-devel
18:38 < helmut> multiarch for its own sake is not overly useful. it should
always be about the use cases of multiarch.
18:38 < _rene_> true
18:39 < _rene_> (that counts for you, there seems to be people who bug people
with "convert to multiarch" bugs just for multiarch sake)
18:39 < gvaillan> hi all, everyone knows why when I run `sudo sbuild-update
--keygen` it fails with the following error `gpg: Invalid
option "--pinentry-mode"`?
18:40 < _rene_> (as the original reporter to the bug you replied to)
[...]
19:42 < helmut> _rene_: so does that mean you an move at least the .pc file or
do I need to get a discussion with Mithrandir going?
19:47 < _rene_> helmut: I won't do it half, so no, I will not move the pc file
only
[...]
20:53 < _rene_> helmut: yes, and I have clearly said that not doing multiarch
in my packages _was intentional_
20:54 < _rene_> helmut: so that won't happen. guess why these bugs are all
still open? :)
20:55 < _rene_> I don't mind involving the tech-ctte here, this will at least
give a clear decision whether multiarch is mandatory or not
20:56 < _rene_> but I've not done it for the last time (except where I forgot
to change it and it was "too late" or where it was in a
multi-arch dep chain...) and won't immediately change that.
even new packages from me get --libdir=/usr/lib or similar
[...]
21:51 < _rene_> helmut: "Failure to do so will result in me NMUing your package
21:51 < _rene_> "
21:51 < _rene_> helmut: threatening does not help. in any way.
21:52 < _rene_> helmut: as I said in my reply to that bug: any NMU doing this
will be superseded by a MU reverting this
21:54 < helmut> _rene_: just make it work and I'll stop bothering you
21:54 < helmut> (where it == cross building rdeps)
21:57 < _rene_> not my "job"
21:57 < _rene_> the package is correct
21:57 < _rene_> it installs a pkg-config file which is found in normal builds.
21:57 < _rene_> end of the story
22:00 < _rene_> just posting the IRC log to the bug
Regards,
Rene
