On Tue, Dec 23, 2008 at 09:11:46AM +1100, Mark Goodwin wrote:
> I unsure, but don't think a group writable repository would work too well.
> Nathan doesn't push directly into the official pcp git tree (which is at
> git://oss/pcp/pcp.git) but rather pushes to his own git tree (which is
> at git://oss/nathans/pcp.git). SGI review and pull those commits and push
> them into the official tree (via an internal staging tree using a hook
> for the ptools back-end, but that's transparent to oss users). Nathan also
> reviews patches from others and pulls their patches into his tree, and
> regularly re-syncs with the official tree. git works it all out just fine.
> This is fairly new, but seems to work rather well since it supports
> effective collaboration for trusted contributors without SGI becoming
> a patch acceptance bottleneck - sort of a multiple maintainership.
I've set up a kernel.org hierachy where can have group commits. We can
try for a while to have sgi pull from it.
>> One thing we were discussing is if it's really a good idea to have all
>> these together.
> It would certainly help SGI if the directory structure for the proposed
> xfs-cmds tree remained the same as it is in ptools at the moment. I
> guess we could consider splitting each xfs-cmds directory into separate
> repositories, but then building it all togetheer would be a pain. It could
> certainly make sense to split off xfstests into it's own tree since it's
> not part of the xfs-cmds build.
Ok, sounds fair to keep it like that for now.