Bug#725143: Bug#728521: ogre-1.9 / gcc-4.8

[Manuel A. Fernandez Montecelo]

Hi,

2013/11/30 Rebecca N. Palmer <r.palmer at bham.ac.uk>:
> Control: severity 728521 normal
> Control: merge 725143 728521

I think that your attempt to merge the bugs failed.  It's OK to keep
them separate because they are slightly different, so it doesn't
matter.


> It's now official that gcc 4.8 will be required in jessie [1], so this is
> now not an RC (and could even be closed altogether, though that isn't my
> decision to make; the current non-working attempts to select gcc-4.8 are
> ugly, but harmless wherever 4.8+ is the default).
>
> s390x are currently switching to 4.8 [2], so another try once they have
> should fix the "out of date on s390x" testing blocker.
>
> [1] http://lists.debian.org/debian-devel-announce/2013/11/msg00007.html
> [2] http://lists.debian.org/debian-devel/2013/11/msg00449.html

It's been built successfully in s390x and most other arches (at least
the ones where it was present before with src:ogre-1.9) so I think
that it will migrate to testing by the end of the year.

I think that the best solution is to let things be and let the arches
catch up with GCC 4.8, I've already been similar reports affecting
packages much more important than OGRE so I think that they will
forced to move on quickly.

I am not keen on depending on two different versions of a compiler for
general as well as specific reasons for this case:

- the fact that 4.8 it's not the default in some arches it indicates
already some kind of problem with that version, so if it remains
unfixed and cannot become the default, maybe the package hits
precisely the problems that make GCC not be the default (OGRE's source
is quite big in terms of LoC, memory needed for compilation, storage,
etc; and takes many hours to compile in most of the arches)

- 4.6 is comparatively quite old, especially for C++ (OGRE's
implementation language) and the latest revisions of the language. Any
minor revision .bugfix could introduce syntax that old compilers choke
with.

- The solution of "gcc-default (>= 4.8) | gcc-4.8" would have been
nice if not because of the problem with buildds (I thought that it
would already have been fixed, it's an old problem).  But the contrary
is not so nice because when GCC 4.9 becomes the default (it can happen
before summer, unless Release Managers want 4.8 for jessie) the
package will need an update just because of its dependencies; while
there's sometimes more than a year [1] between updates to OGRE package
and for good reasons (need to go through NEW queue due to changes in
names of binary packages; tax on the infrastructure; etc).

[1]
[2013-09-04] Accepted 1.8.0+dfsg1-5 in unstable (low) (Manuel A.
Fernandez Montecelo)
[2013-08-01] Accepted 1.8.0+dfsg1-4 in unstable (low) (Manuel A.
Fernandez Montecelo)


So as I said, I think that it's better let things be and the problem
will solve itself plenty of time ahead of the freeze.  Reverse-depends
can ask to remove old versions of packages from testing in problematic
arches to get their new version to migrate.  Virtually nobody will be
using OGRE or their packages in those arches anyway.


Cheers.
-- 
Manuel A. Fernandez Montecelo <manuel.montezelo at gmail.com>