On 02/08/14 21:59, John Paul Adrian Glaubitz wrote: > Thanks a lot for the detailed explanation. I was never aware that the > use of pkg-config for the headers was mandatory for GLib and libdbus, > but it actually makes a lot of sense and seems a very clean solution > since it forces you to think about the proper inclusion of the header > files yourself.
It doesn't force you to think about the include path (in a way, part of the point of it is to *not* have to think about the include path), but it does force you to declare which API version you're trying to use (indirectly, by your choice of .pc file). GLib and libdbus were maybe bad examples: Gtk is a better example, since it has two widely-used incompatible API versions, with pkg-config names gtk+-2.0 and gtk+-3.0, each of which expects you to #include <gtk/gtk.h> (but you can have both installed in parallel, because the -I options passed to the C compiler are different, and you also end up linking a differently-named library). <http://ometer.com/parallel.html> was written by a GNOME developer during the transition from Gtk 1.2 to Gtk 2, and explains the general design, which was re-used for the transition from Gtk 2 to Gtk 3. I believe the Qt/KDE world has taken a similar approach in Qt 3/4/5 and KDE 3/4. > That looks very reasonable to me and I would actually favor your > solution over Mateusz' suggested patch. I'd be ok to have it uploaded > like this, so please go ahead. Then I could finally work on Mateusz' > deadbeef package again which is still on mentors waiting to get > reviewed and sponsored. If you are an uploading DD and you actively use libzip, I would prefer you to do an NMU based on my suggested patch: it seems you know how to test it, and I don't really. I found this bug on the RC-bug list, and I have no particular interest in libzip myself. S -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org