Pavel Zhukov wrote:
I've pushed aws-3.2.0, matreshka-0.7.0 and gnatcoll-2014 to
rawhide
yesterday. Our internal project requires new aws and gnatcoll
(because of bug with int overflow in json parsing) so I'm thinking
about so version bump in F22 as well.
Is it possible to backport the fix?
When thinking about changing a soname in a stable release, keep in mind
that you don't know how many users there are out there who use the
library in their internal programs. Their program will fail to start
the next time they start it. Presumably they can recompile it, but
that's no fun if they were just restarting a daemon to pick up a
configuration change.
Gnatcoll was compilled without gtkada support (it has to be updated
as
well).
We have GTKada 3.8.2 and I don't see a newer release than that from
Adacore. Does it work better if you require GtkAda3-devel instead of
GtkAda-devel?
Feel free to change with_gktada once gtkada package will be
updated
Me? I can't make any changes to Gnatcoll unless you give me commit
access.
The only problem we have is gprbuild. It doesn't work correctly
in
current state because of "builtin" sources. They're sources of
gcc-4.X. I've opened the bug against gcc to package gnat-sources as
well as we do for ocaml sources (BZ#722732) but Jakub is not happy
with the request :( for now gprbuild ships content of gcc/ada in
%{source2} which is dirty hack and I don't like it at all.
Yes, that's a bit ugly. The real solution would of course be to have GCC
upstream make those source files into some kind of library. That's
probably not going to happen in Fedora until it happens upstream.
I think it would be somewhat cleaner to include the whole GCC source
tarball as Source2 and set GNAT_SOURCE_DIR to the gcc/ada directory in
the GCC source tree. That would make it much easier to verify that the
sources are genuine. Minor changes like the one you made in
sem_util.adb could be done with patches as usual.
Björn Persson