Should MariaDB touch my.cnf in %post?

[Miloslav Trmač]

On Mon, Feb 18, 2013 at 11:05 AM, Reindl Harald <h.reindl at thelounge.net> wrote:
> Am 18.02.2013 10:58, schrieb Honza Horak:
>> It seems the words could have been chosen less absolutely (changed a little bit at [1]), but in case you don't
>> agree with the official MariaDB statement, that "For all practical purposes, MariaDB is a binary drop in
>> replacement of the same MySQL version", "MySQL 5.5 will be compatible with MariaDB 5.5" and "The shared client
>> library is binary compatible with MySQL's client library." [2] -- I'd like to see concrete examples where this
>> incompatibility is broken, so that we can document it at least or better to fix it asap.
>
> as said - this is thin ice
> remember MySQL 5.5.10
>
> you can not expect real binary compatibility over the
> whole release cycle and if it slightly breaks it starts
> to get funny in context having both in Fedora and depend
> on exact obsoletes / provides / dependencies

"Real compatibility over the whole release cycle" is a higher bar than
we need to achieve.

What we actually need is much weaker, namely API (not ABI)
compatibility between F18 and F19, and perhaps we might consider ABI
compatibility for non-Fedora pre-compiled binaries.

We don't need ABI compatibility between F18 and F19 for Fedora
packages because there will be a mass rebuild.

We don't need ABI compatibility between the two implementations "over
the release cycle" because once the client libraries are switched to
the MariaDB implementation, ABI changes in the MySQL implementation no
longer concern Fedora packages.  Non-Fedora pre-compiled binaries
would be affected by the divergence, but this divergence is no worse
than the divergence in ABIs between F18/F19/F20 or between the
corresponding ABIs of MySQL client library versions.
    Mirek