CVS commit messages and backpatching

Bruce Momjian wrote:

FYI, in going through the release notes, I would like to remind
committers that it is important to mention if the commit was
backpatched to any earlier release.

Backpatches usually happen after the commit to the head branch has been
made, so the information is not available at that time.

--
Peter Eisentraut
http://developer.postgresql.org/~petere/

Peter Eisentraut wrote:

Bruce Momjian wrote:

FYI, in going through the release notes, I would like to remind
committers that it is important to mention if the commit was
backpatched to any earlier release.

Backpatches usually happen after the commit to the head branch has been
made, so the information is not available at that time.

Well, I have a script where I create a single commit message that I use
to apply to multiple branches. Usually you know at the time of commit
to HEAD that you are going to backpatch. And in such cases, it is
helpful to mention it as part of the HEAD commit.

--
Bruce Momjian bruce@momjian.us
EnterpriseDB http://www.enterprisedb.com

+ If your life is a hard drive, Christ can be your backup. +

Bruce Momjian <bruce@momjian.us> writes:

FYI, in going through the release notes, I would like to remind
committers that it is important to mention if the commit was backpatched
to any earlier release. Not only is this valuable for making the
release notes, but it also helps people looking at the commit message.

With the standard output from cvs2cl, this is pretty obvious anyway, no?
I see entries like

2006-08-29 09:39 teodor

* contrib/tsearch2/: tsvector.c (REL8_1_STABLE), tsvector.c: Remove
pos comparison in silly_cmp_tsvector(): it is not a semantically
significant

so it seems to me that explicit mention of back-patching is mostly
redundant. (Of course, this requires the committer to commit all the
branches at about the same time, which I make an effort to do precisely
so that the cvs log looks nice. If some time elapses between patching
and back-patching then a mention in the commit message is definitely
needed.)

regards, tom lane

Tom Lane wrote:

Bruce Momjian <bruce@momjian.us> writes:

FYI, in going through the release notes, I would like to remind
committers that it is important to mention if the commit was backpatched
to any earlier release. Not only is this valuable for making the
release notes, but it also helps people looking at the commit message.

With the standard output from cvs2cl, this is pretty obvious anyway, no?
I see entries like

2006-08-29 09:39 teodor

* contrib/tsearch2/: tsvector.c (REL8_1_STABLE), tsvector.c: Remove
pos comparison in silly_cmp_tsvector(): it is not a semantically
significant

so it seems to me that explicit mention of back-patching is mostly
redundant. (Of course, this requires the committer to commit all the
branches at about the same time, which I make an effort to do precisely
so that the cvs log looks nice. If some time elapses between patching
and back-patching then a mention in the commit message is definitely
needed.)

I pull activity only from HEAD, so I do not see that tag. In fact, I
use our src/tools/pgcvslog rather than cvslog.

--
Bruce Momjian bruce@momjian.us
EnterpriseDB http://www.enterprisedb.com

+ If your life is a hard drive, Christ can be your backup. +

Bruce Momjian <bruce@momjian.us> writes:

I pull activity only from HEAD, so I do not see that tag. In fact, I
use our src/tools/pgcvslog rather than cvslog.

Not sure why we are maintaining our own script when there are much
better things out there:
http://freshmeat.net/projects/cvs2cl.pl/

regards, tom lane

Tom Lane wrote:

Bruce Momjian <bruce@momjian.us> writes:

I pull activity only from HEAD, so I do not see that tag. In fact, I
use our src/tools/pgcvslog rather than cvslog.

Not sure why we are maintaining our own script when there are much
better things out there:
http://freshmeat.net/projects/cvs2cl.pl/

Well, my script produces output that is closer to what I need to create
the release notes.

--
Bruce Momjian bruce@momjian.us
EnterpriseDB http://www.enterprisedb.com

+ If your life is a hard drive, Christ can be your backup. +

Bruce Momjian wrote:

Tom Lane wrote:

Bruce Momjian <bruce@momjian.us> writes:

I pull activity only from HEAD, so I do not see that tag. In fact, I
use our src/tools/pgcvslog rather than cvslog.

Not sure why we are maintaining our own script when there are much
better things out there:
http://freshmeat.net/projects/cvs2cl.pl/

Well, my script produces output that is closer to what I need to create
the release notes.

If there are procedures, please document them. Nobody ever told me much
when I was given committer status, and I just did what it looked like you
guys did, and no doubt made some mistakes.

cheers

andrew

Andrew Dunstan wrote:

Bruce Momjian wrote:

Tom Lane wrote:

Bruce Momjian <bruce@momjian.us> writes:

I pull activity only from HEAD, so I do not see that tag. In fact, I
use our src/tools/pgcvslog rather than cvslog.

Not sure why we are maintaining our own script when there are much
better things out there:
http://freshmeat.net/projects/cvs2cl.pl/

Well, my script produces output that is closer to what I need to create
the release notes.

If there are procedures, please document them. Nobody ever told me much
when I was given committer status, and I just did what it looked like you
guys did, and no doubt made some mistakes.

I guess the question is whether it is possible using cvs2cl to show only
HEAD, and then show if the same commit message also appears in the most
recent back branch. And will that always work reliably?

--
Bruce Momjian bruce@momjian.us
EnterpriseDB http://www.enterprisedb.com

+ If your life is a hard drive, Christ can be your backup. +

Bruce Momjian <bruce@momjian.us> writes:

I guess the question is whether it is possible using cvs2cl to show only
HEAD, and then show if the same commit message also appears in the most
recent back branch.

It's so rare that we make a back-branch patch without a corresponding
HEAD patch that I'm not clear why you are concerned about showing only
HEAD for this purpose. I've always found that cvs2cl's behavior shows
me exactly what I want to know for CVS log searching (and that includes
release note making). If the output format isn't quite what you want,
maybe you could turn on its XML-output option and then munge that into
HTML. I've never had occasion to play with the XML-format option myself.

regards, tom lane