> > log, and removed the file format change to store the author (since we get
> > it elsewhere now).
>
> Ah. Very good. Yes, this patch is much cleaner. However, I haven't committed
> it in its exact form, so I'll do a detailed review of this to let you know
> what I kept, what I skipped, and what I did a bit differently.
>
> The changes are pretty big, but one fact remains: you led the way on getting
> the commits to occur, and I'd like to give a big THANKS! for that.
>
> As a test case, I converted the mod_dav CVS repository in under 10 minutes
> (480 distinct commits). It looks like there is a still some minor issues
> with determining the groups of a commit, but that should be fixable. The
> bulk of the logic to actually commit a transaction appears to be working.
>
> [ I still want to write a script to extract each SVN commit, extract each
> CVS tree at a date, and compare the two trees ]
>
> [ to fix the commit groupings, I'm going to ensure each commit only includes
> a file once; if a file occurs a second time, then we will need to slice
> the group into one or more groups. ]
This, btw, won't fix the author/log problem.
You need a seperator in between log and author before digesting them.
Otherwise, things like spelling mistakes and accidental typos in logs, as
well as the usernames you've used in a cvs repository, could become
significant.
IE it'll give:
Log="This patch is good.d" Author: "berlin"
and
Log="This patch is good." Author: "dberlin"
The same ID.
While this example may be a bit silly, I can think of actually
plausible situations where this type of things happens. In fact, it
hit me tonight while converting a repo.
Just throw a separator in between (log + author), and fixing commit
grouping should make the problem go away.
--Dan
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 21 14:37:06 2006