Ken Erickson <ken.erickson at sun.com> writes:
> I think there are (at least) 2 classes of hooks that we need. The
> basic ones are those that
> facilitate daily gate operation - notification of commits to everyone,
> automated way
> of adding new "gatelings" like the add-gateling script does now, and
> notification
> of a successful commit to the committer (like pbconfirm does now).
notification is handled by existing stuff on opensolaris.org Hg also
has an extension to provide notification.
One or the other of those cover the first thing.
add-gateling, at least opensolarisly, is the wrong thing to do. I
leave it to your judgement as to whether you'd like it to continue to
be there while the gate is internal.
("everyone who brings over" is a very large set, when you consider
opensolaris.)
pbconfirm is probably slightly complex, in the same way add-gateling
is now, you have to figure out a real email address for the user. It
should be implemented in much the same way as either our, or
mercurial's, notification support, I think.
> Another set of hooks which, presumably, will be different between SFW and ON
> are things that check a commit before letting it happen. In SFW land,
> we are pretty
> liberal, and we don't enforce much "correctness checking" - not that
> we couldn't, but
> we don't yet. In ON-land, though, there is much more of a "tight-ass"
> factor to deal with,
> so I assume a bunch of scripts will be run on the files to be
> committed before the commit
> occurs.
Yes, this has been our plan all along. This is why the "Checks" code
we have is modular, such that all checking can use the *same* checks.
I hadn't been considering those things as part of the gk bits.
> I am most interested in group #1 above, since that's what we need to
> do to get SFW going at a minimum.
>
SFW does need a subset of correctness checks, but as above, I hadn't
considered those as if they were gk bits, exactly, so I think I agree
with that, yes.
-- Rich
