Re: AW: Release timestamps for archaic tarballs

[Eric S. R. <es...@th...>]

Daniel J Sebald <dan...@ie...>:
> This may be where Eric's approach has an advantage, that rather than the
> diff hunk it sounds like he's pulling out the actual snapshot of the
> ChangeLog at the time the changeset was made.  That might sweep some more
> changesets under the same author.

That is correct.  My code walks the commit list looking at the actual
blobs associated with each changeset.  And those are whole-file
copies, not diffs.

This should not fail except in a couple of unlikely cases

(1) A malformed ChangeLog entry header (I've found two of those).

(2) Somebody tweaks the middle of a ChangeLog file (e.g. to fix a typo) without
    adding a new top entry and it is *not* a changeset modifying only
    ChangeLogs.

(3) Too-small fuzz (time window for CVS commits to be considered close
    enough to coalesce into a changeset) results in a changest being
    broken into pieces. It's hard to know how often this happens, but
    I think it's pretty rare.

One of my metrics for a conversion that was worth doing is when I find a good
reason to write and field-test a new surgical primitive.  GNUPLOT has, a bit
to my surprise, been interesting twice:  one new primitive is the Changelog
miner, the other is a command to incorporate tarballs as new commits.
-- 
		<a href="http://www.catb.org/~esr/">Eric S. Raymond</a>

My work is funded by the Internet Civil Engineering Institute: https://icei.org
Please visit their site and donate: the civilization you save might be your own.