Wondering about Gentoo doing automated builds (aka tinderboxing which
flameeyes and bonsaikitten were up to) as part of its process. Seems to be
a move to git, and git-sha IDs came up wrt caches on dev. (If we had those
as a digest type, they'd effectively be free to 'generate' since they'd
already be in place.) Instead of splitting at branch level, if we split at
repo level, so we had a QA repo where the mdata branch lived (if we did
want to split there as well) that'd be an ideal place to base our tinderbox
work from. (At server side the bare clone would take very little resource,
and we could pull into a fresh clone at the build machine if it were a
different host.)

It seems silly to switch to git for development work and not take advantage
of cheap clones to inculcate a slightly more professional process, at the
time when it's easiest to do so. The advantage would be that we'd catch
more errors before they went to users, and we'd have a point where we can
start to do other more interesting things, eg around binpkgs and
semi-frozen trees.

By having it as part of Gentoo's process, there would be less need for
externals to devote hours of CPU resource to building the whole tree, and
bugzilla wouldn't get bombed with the results; developers would get a
pingback (on IRC if there, by email if not) to fix their commit as it's not
made it to the tree yet. Gentoo would not need to build whole tree over and
over either, just the new stuff as and when it comes in.

Brought this up here as it's more process than ebuild development, and can
be moved to dev easily enough, should it warrant it.

-- 
#friendly-coders -- We're friendly but we're not /that/ friendly ;-)