Version in mysql client reported incorrectly in deb packages since big packaging change

Bug #1319670 reported by Roel Van de Paar on 2014-05-15
This bug affects 2 people
Affects Status Importance Assigned to Milestone
Percona Server moved to
Fix Released
Tomislav Plavcic
Fix Released
Tomislav Plavcic
Fix Released
Tomislav Plavcic
Fix Released
Tomislav Plavcic

Bug Description

When starting mysql client:

Server version: 5.6.17-65.0-587.trusty-log (Ubuntu)
Server version: 5.5.36-34.0-632.wheezy (Debian) wheezy

Same output seen in \s status command

But 5.5.35 is mostly fine (thought the release number is repeated and there is no revid):
Server version: 5.5.35-33.0 Percona Server (GPL), Release 33.0

To compare, this is RPM:
Server version: 5.5.37-35.0 Percona Server (GPL), Release rel35.0, Revision 657

And this is a Precise output:
5.6.16-64.2-tokudb-log Percona Server (GPL), Release 64.2-tokudb-7.1.5, Revision 538 for precise

So this regression was introduced at the time of Stewart's large MP.

The build scripts may have been changed at that point to make it possible to release with the new layout.

This is only for deb packages, RPM and Tarball are unaffected.

Related branches

tags: added: pkg
Roel Van de Paar (roel11) wrote :

Let's standardize this across rpm/deb/tar. How about:

 Server version: 5.x.xx-xx.x Percona Server (GPL), Release xx.x, Revision xxx

Does everyone agree on this? Potentially we could drop the Release xx.x, Revision xxx bit and instead have:

 Server version: 5.x.xx-xx.x-xxx Percona Server (GPL)

Anything wrong with this?

Btw, the TokuDB bits as I understand it can be dropped, see mailing list (and PS would be build without TokuDB engine anyways)

description: updated

The outdated version needs to be fixed as a high priority and, depending on the cause, necessary steps added to the release checklist to avoid this in the future.

Re. the unification, I wouldn't change any string formats at all, as these might be parsed by tools etc. The only thing that I think needs unifying is ensuring that version-affecting CMake options, e.g. COMPILATION_COMMENT, are set consistently across all packages.

Roel Van de Paar (roel11) wrote :

Release checklist already updated.

Roel Van de Paar (roel11) wrote :

Comment #3 was referring to manual release checklist done after packages are build. However, this may also need a change in the RM steps. Though already well covered (see below), depending on what we find, we may need to adjust these steps to be a bit more detailed (check X in file Y):

 [Builder] Check version number is correct: source, packaging
 [Builder] Check version number is correct: XtraDB version in univ.i

Stewart Smith (stewart) wrote :

Not sure why the debian version name is getting into the server version.... perhaps this is some side effect of how the builds are being done... IIRC the server version string should come straight from the magic VERSION file in the source tree, with additional version bits for debian packaging, but those ~wheezy bits shouldn't be shown in server output.

Something odd somewhere... certainly a bit of version number stuff changed with my giant MP, but the full version string should just come from the fields in VERSION.

Tomislav Plavcic (tplavcic) wrote :

Maybe I'm wrong but distribution part is put into MYSQL_SERVER_SUFFIX cmake option so it gets appended to server version - I think this just needs to be removed but will test it.

The thing I would like to know is the format of compilation comment. Is this one ok (without revision)?
Percona Server (GPL), Release xx.x
so the whole thing would look like this for all rpm/deb/tar:
Server version: 5.5.35-33.0 Percona Server (GPL), Release 33.0

Revision is currently passed to rpm spec from jenkins, and for debian rules it is not passed - so need to know if we need this part or not - it would be cleaner if it's not needed.

Roel Van de Paar (roel11) wrote :

See PS-155, Bug 1327817

Percona now uses JIRA for bug reports so this bug report is migrated to:

To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Duplicates of this bug

Other bug subscribers