Re: missing tags

Magnus Hagander <magnus(at)hagander(dot)net> writes:
> On Sat, Oct 2, 2010 at 13:36, Andrew Dunstan <andrew(at)dunslane(dot)net> wrote:
>> There are commit messages from about 22 hours ago that say that the

> This was intentional - to wait with the tags until the tarballs have
> been verified *and published*, so we don't end up having to move the
> tags later if we find a last-minute problem.

Yeah. Given that moving/removing tags in Git is next to impossible,
we decided not to follow our old practice of applying tags immediately
after the version-stamping commits anymore. We'll tag later once we're
sure a respin isn't going to happen. It's easy and reliable to place a
tag on a past commit (unlike in CVS), so that should work fine. This
decision already paid benefits: if Marc had already pushed a 9.0.1 tag
when he discovered the docs didn't build, we'd have had a problem.

Just for the record, I believe this is where the tags will go, barring
we find another need to respin:

git tag REL9_0_1 a0ccae7ed41201251c0da51a91b84eb9d13c4ab1
git tag REL8_4_5 2f76a4b5a6bcefa03f5a4d377cfb449fd5d95185
git tag REL8_3_12 e32229adfa0769281b0650c8f999d54079c97661
git tag REL8_2_18 61318f38c0999e828d34230c3ea57eb24c49ceba
git tag REL8_1_22 d6d7926cd198cd8d42b0f2d40cbd8f876be9f278
git tag REL8_0_26 c689dacd6deba81c742ec35b2b792fba47db1ffc
git tag REL7_4_30 fd7fdbc88be20f5fc8789254b18902116f88b5fe

Hopefully that agrees with Marc's notes.

> I think the confusion is from the use of "tagged" in the commit
> message

Possibly Marc should adopt the habit of making the commit messages read
like "Stamp 9.0.2", rather than "Tag".

regards, tom lane