[Evolution] Bug#792494: evolution-data-server fails to build with GCC 5 from experimental

[Simon McVittie]

On Wed, 15 Jul 2015 at 14:05:22 +0200, Matthias Klose wrote:
> patch at
> https://launchpad.net/~ci-train-ppa-service/+archive/ubuntu/landing-016/+files/evolution-data-server_3.12.11-0ubuntu3~gcc5.1_3.12.11-0ubuntu3.diff.gz

I've adapted that patch on the assumption that e-d-s upstream do not consider
the offending symbols to be ABI (they're C++ template instantiations from
a private module).

> also:
> library transition may be needed when GCC 5 is the default

I think this is a false positive because those symbols aren't API.

> The libstdc++6 transition will be a large one, and it will come with a
> lot of pain.  Please help it by preparing the follow-up transitions.

To reduce the pain slightly, the attached patch also disables
libphonenumber support on Linux for the moment; according to codesearch,
only syncevolution uses it, and syncevolution seems to be designed to
cope gracefully with its absence.

This would let the buildds start work on libreoffice as soon as e-d-s
has built; this seems like a big win, because libreoffice takes hours
to build, and has had a pair of successful binNMUs on kFreeBSD (where
libphonenumber is already disabled).

I have not been able to test the patched e-d-s yet, since upgrading
my system to the g++-5 ABI still has too much collateral damage (notably
including libreoffice becoming uninstallable). However, perhaps it's
worth uploading anyway, to keep the transition moving?

    S