Hi everyone,
Sorry for coming into this discussion late, but having a thread with
58 messages in it is a bit intimidating!
In 1989 I was actually a full-time "build sheriff" for 9 months at
Alias. The terms were a bit different - the software repository was
called "the repo" and the build sheriff was called the "repo
man" (cue the Iggy Pop music). Over the years the system was
automated, made more complex and the current solution is to disallow
commits that fail a specific set of criteria, which varied, depending
on where the commit touched -- compile/link/test for code, spelling/
html validation for doc. This causes the committer to fix the problem
and re-submit the commit. While waiting for the confirmation email,
the committer is still able to work, based on their as-yet-
unconfirmed commit.
An obvious drawback to this might be blocking someone else who is
waiting for the first person's work to be committed. However more
than one user can share a commit, which is called a "changeset"
before it is committed. I'm not sure if svn can do this. Based on my
quick perusal, it can't. Failing this, it would make multi-user
dependant commits a bit tricky in terms of timing.
Would a commit-refusal system be feasible?
Reid
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Open Source Applications Foundation "Dev" mailing list
http://lists.osafoundation.org/mailman/listinfo/dev
