package mariadb-client-5.5 '/usr/bin/innochecksum' needs to break/replace mysql-server-core-5.6 OR fix the binary/manpage mismatch in mysql

This bug is because in mysql the innochecksum binary and man page are shipped in different packages (client vs server). In MariaDB the innochecksum binary and manpage are both part of the client package.

We need to either fix MySQL packaging or add mysql-server-5.6 into the breaks/replaces of mariadb-client-5.5 (and -10.0, which should be part of vivid instead of 5.5), like in done in commit http://anonscm.debian.org/cgit/pkg-mysql/mariadb-5.5.git/commit/?id=52ceea23d59849801cd6359cac31804cab8153cb when the innochecksum binary/manpage mismatch was fixed in MariaDB.

If you need a quick fix, then simply run 'sudo apt-get remove "mysql-*"' and then 'sudo apt-get install mariadb-server' to bypass this breaks/replaces issue.

Otto,

I moved the innochecksum binary to the server package, because it clashed in a similar way to percona-server, and we concluded on IRC that it made more sense to be in the server package (as it requires server data files to operate).

I didn't move the manpage, but I should do that. I also didn't consider that there might be a clash with mariadb also - sorry about that.

Can we agree on the right place for innochecksum to live first, and then we can make sure that all variants do the same thing? How do you feel about moving it to the mariadb server package?

2015-02-13 15:13 GMT+02:00 Robie Basak <email address hidden>:
> Can we agree on the right place for innochecksum to live first, and then
> we can make sure that all variants do the same thing? How do you feel
> about moving it to the mariadb server package?

OK, let's do that

Robie Basak, you have triaged this bug without assigning priority to it.

Let's hold this in vivid-proposed until I have resolved it, to avoid users breaking when mysql 5.6.23 hits vivid.

Otto, to confirm, what I've understood from this is to move /usr/bin/innochecksum and /usr/share/man/man1/innochecksum1 from mariadb-client-10.0 to mariadb-server-core-10.0. I'll push this to Debian VCS for MariaDB now.

Then I can add a Breaks/Replaces on mysql-server-core-5.6 against mariadb-client-10.0 (< before_above_change). Since the mariadb and mysql -server-core packages cannot be unpacked concurrently already, I think this should be all that is needed.

Oh - this will fix mariadb-10.0. I have no plans to fix mariadb-5.5, since this is to be removed from Vivid this cycle.

mariadb-10.0 fix committed to Debian VCS.

mysql-5.6 fix committed to Debian VCS, and uploaded to vivid-proposed. Now being held due to dep8 failures due to bug 1423279, that's being worked on, and bug 1417328, that should be fixed as soon as I've got percona-xtradb-cluster-5.6 uploaded.

This bug was fixed in the package mysql-5.6 - 5.6.23-1~exp1~ubuntu3

---------------
mysql-5.6 (5.6.23-1~exp1~ubuntu3) vivid; urgency=medium

  * Call dh "--with systemd" to enable systemd unit.
 -- Martin Pitt <email address hidden> Mon, 09 Mar 2015 09:38:35 +0100

This bug was fixed in the package mariadb-10.0 - 10.0.16-2~exp1~ubuntu1

---------------
mariadb-10.0 (10.0.16-2~exp1~ubuntu1) vivid; urgency=medium

  * Merge from Debian VCS commit e73d494. Remaining changes:
    - Build without TokuDB since it fails tests.
    - Depend on mysql-common (>= 5.6.22-1~) to ensure that the new
      my.cnf symlink wrapper is available during upgrade.
    - Add versioned mariadb-common dependency to mariadb-server-10.0 so
      it doesn't fulfil the dependency using an older mariadb-common
      that doesn't require the newer mysql-common with the my.cnf
      symlink support.
  * Drop changes (applied in Debian):
    - Fix typo in mariadb-server-10.0.postinst.
    - Include change by Otto found in ppa:mysql-ubuntu/mariadb-10.0 but
      not yet in VCS: include "/etc/mysql/conf.d/*.cnf" for global options.

mariadb-10.0 (10.0.16-2~exp1) UNRELEASED; urgency=medium

  * Move innochecksum back to mariadb-server-core-10.0 to align with other
    variants (LP: #1421520).
  * Fix typo in mariadb-server-10.0.postinst.
  * Fix typo in postinst mktemp call (LP: #1420831).
 -- Robie Basak <email address hidden> Tue, 17 Feb 2015 19:19:05 +0000