When we do a final release, as we're about to for 3.5, we need to tag the latest _RELEASE changeset with a final tag as well. Eg, FIREFOX_3_5rc3_RELEASE and FIREFOX_3_5_RELEASE will point to the same changeset soon. We can do this with a bit of hackery currently but because we have to tag en-US + all locales we really need to able to do this in a better way.

Mass move of bugs from Release Engineering:Future -> Release Engineering. See http://coop.deadsquid.com/2010/02/kiss-the-future-goodbye/ for more details.

This should be made easier when tagging is moved to an external script, in bug 508896.

This didn't turn out as clean as I wanted, but it does the job. We don't use a plain two octet version for anything except final releases at this time (RCs use version=4.0rc1, appVersion=4.0), so I believe this is safe. When we come to final tagging of 4.0, I expect us to be using the following in the config: sourceRepoRevision = (changeset of final RC) relbranchOverride = (relbranch of final RC) baseTag = FIREFOX_4_0 version = 4.0 appVersion = 4.0 skip_build = 1 ...and then kick off the automation as normal, which will then run tag, source, and xulrunner_source. I also toyed around with having a flag passed that controlled final tagging behaviour, but I didn't go with that for a couple of reasons. We'd have to either: - Decide in Buildbot that we want to pass it (forces us to reconfig for final tagging) or decide in tagging.sh (which would be messy, and not the right place IMHO) - Use a separate Builder for it, which seems unnecessarily burdensome, and would require us to fire Source separately or have a final-release specific source builder.

Comment on attachment 493416 [details] [diff] [review] add final tagging support to tag-release.py >+# If version is two octects with no trailing specifiers like "rc", we >+# consider it a "final" release for which we only create a _RELEASE tag. >+FINAL_RELEASE_REGEX = "^\d+\.\d+$" Small typo of "octet" there....and I'm not sure that octet is the right term for these? >+def isFinalRelease(version): >+ # "not not" is used to avoid returning a match object >+ return not not re.match(FINAL_RELEASE_REGEX, version) Would bool(re.match(...)) work? r+ either way here

Carrying forward r+

Comment on attachment 493757 [details] [diff] [review] address catlee's review comments changeset: 1327:a13ca99a1384