rawhide report: 20070912 changes
Nalin Dahyabhai
nalin at redhat.com
Thu Sep 13 16:28:06 UTC 2007
On Thu, Sep 13, 2007 at 04:10:34PM +0000, Kevin Kofler wrote:
> Ralf Corsepius <rc040203 <at> freenet.de> writes:
> > > At one point in time, all
> > > the options had to be supplied to the "configure" script.
> > This still applies, this is what "packaging is about".
>
> You're forgetting that not everyone who compiles software is a packager. Having
> to set CFLAGS and LIBS by hand might be acceptable to you as a packager, but
> most upstream projects will not consider this an acceptable solution, because
> they also have to support compilation from a tarball and want to make that as
> simple as possible.
>
> Now, of course, adding the .pc file in Fedora only won't help users on other
> distributions, but that's why you as a maintainer are expected to talk to
> upstream about such issues instead of sticking your head in the sand.
Forgive me for wading in here, but upstream *has* to be where .pc files
show up, and if they don't show up there, we absolutely shouldn't be
adding them to binary packages. I believe this very strongly.
Including a .pc file in a -devel package suggests, to maintainers of
projects which use that -devel package, that the .pc file can be
depended upon to always be there when the library is, that it's safe to
have configure scripts depend on their presence.
If I were a maintainer of a package which depended on a .pc file, and I
started getting reports from people who couldn't install my package
because they built a depended-upon package from source (because they're
on another operating system or Linux flavor), and I then tracked the
root cause down to my dependence on a .pc file which isn't available
everywhere, I'd feel betrayed.
Cheers,
Nalin
More information about the fedora-devel-list
mailing list