Re: proposed enhancement: better logmsg formatting

[Russ Tremain]

At 3:37 PM -0700 5/10/01, Greg Klanderman wrote:
>>>>>> Russ Tremain <russt@ebay.sun.com> writes:
>
>> Hi Greg -
>> I'm interested in trying out your patch, and
>> have been working in a similar area. I have added a "d"
>> format specifier that will include the revision number of
>> the "dead" revision when a file is deleted.  (We need this
>> to get the rcs log message for the deleted file so we
>> can scan the message for a numerical identifier to hook
>> it to an external "commit notice" facility.)
>
>Hi Russ,
>Hmm, I never tested it, but I would have assumed that the 
>existing {v} specifier would have done this, and if not 
>that it should be fixed to do so.  Or am I confused?

unfortunately, the {v} specifier is overloaded.  it is set
to "NONE" to denote a deleted revision.  And it turns out to
be a non-trivial problem to find the revision number of the
deleted revision, because the {V} specifier will report
the trunk version if the commit is the first on a branch.

This make it difficult to find the commit message for
the delete.

The best solution would be to have an "op" field instead
of overloading {vV}, but that would break compatibility.

>> I'm particularly interested in your "t" specifier, as I was
>> thinking along similar lines.  One question - will {t} retrieve
>> the cvs branch name?  Or will it only retrieve a tag associated
>> with the specific revision that was committed?
>
>> I would like a format specifier that finds the branch tag
>> for a commited file... I was thinking about using "b".
>
>If I understand you correctly, {t} does what you want.  
>You can't commit to a tag that is a revision, only to 
>a tag that is a branch.
>

hmm... didn't think of that!  chicken and egg; can't tag
before you deposit.

this is great, I'm going to try it out and merge in
my {d} changes.

-Russ

>> Thanks very much for sharing your work.
>
>No problem, I'm glad there's interest...
>
>greg