mock configuration

I’ve put together a configuration for mock to allow us to cleanly build SL6 (both i386 and x86_64) packages. So far I’ve only tested that this works on my F13 test machines.

There were a few small issues that are worth noting.

  • There is no ccache package in SL6, this has to come from epel6 instead. This means the ccache plugin must be disabled when epel6 is not available.
  • A local yum repository which provides a “build” package group needs to be created. I grabbed the one from koji for epel6 and installed it locally as part of an RPM. Either the i386 or x86_64 version will do as they are identical. This is used to install the standard set of build-requirements which should be present for all packages.
  • There were some suggestions on the web that mock version 1.1.7 is the minimum for building RHEL6/SL6 packages. I didn’t bother looking into this but just upgraded to the latest version which is in Fedora14 (1.1.8). It certainly appears to be faster if nothing else.
  • It appears to no longer be necessary to use yum exclude options to prevent inclusion of i?86 packages in x86_64 chroots. This seems to be fixed by setting “multilib_policy=best” in the yum.conf section of the mock configuration.

Informatics DICE users can enable the SL6 mock configuration by including the dice/options/mock.h LCFG header. Once updaterpms has been run you just need “om mock start“. Depending on your PAM mock/consolehelper configuration, you may also need to add your username to the mock group, in LCFG this is done something like this:

!auth.grpent_mock       mCONCATQ(":squinney")

I have also updated the mock component information on the LCFG wiki to show the recent changes to the component schema.

This entry was posted in Uncategorized. Bookmark the permalink.