On Tue, 12 Mar 2013 14:33:21 +0100, Honza Horak <hhorak@xxxxxxxxxx> wrote:
On 03/12/2013 02:03 PM, Norvald H. Ryeng wrote:
On Mon, 11 Mar 2013 19:00:29 +0100, Honza Horak <hhorak@xxxxxxxxxx>
wrote:
if I understand it correctly that the problem is caused by conflicting
library names, then it should be solved today (the enhanced package is
already building).
Why not have non-conflicting library names? The APIs are different, so
it makes sense to have both libmysqlclient.so and
libmariadbclient.so. These can co-exist and applications can choose
which library to build against.
That would mean to persuade many depended projects to enhance their
building configuration. Unless these projects start using some
in-compatible features regularly, I don't think it would be worth
changing the library name -- such projects currently don't care if it is
build against mysql or mariadb.
I believe both libraries should be installable and that the upstream
projects and/or the maintainer of the package should choose which library
to use. There are already API differences, so the libraries aren't fully
interchangeable. The libraries are two different implementations of the
same protocol. I don't see why they should use the same soname.
With a parallel installation, a transition from one library to the other
could be done gradually and controlled, one package at a time, with less
risk of breakage.
The solution that's currently implemented (bumping the version of the
original libmysqlclient.so from 18 to 1018) is a hack and not a
long-term solution.
True, I'm expecting MySQL will be rebased to 5.6 and mariadb to 10.0
sooner or later and I don't believe that library versions will remain
the same at that point. What does MySQL upstream actually plan with
library version in 5.6? Is it going to be bumped?
Yes, 5.6 will be in as soon as things settle and we find a workable
solution for having both MariaDB and MySQL in the same distro. We already
have 5.6.10 packaged. We're just waiting for package names, dependencies
and everything else to settle. There's no need to mix yet another variable
into the equation right now. It's complex enough without it.
I'm not sure about the version number. I'll have to check with those
responsible for the the client library.
Regards,
Norvald H. Ryeng
--
devel mailing list
devel@xxxxxxxxxxxxxxxxxxxxxxx
https://admin.fedoraproject.org/mailman/listinfo/devel