On Sat, Jun 07, 2003 at 01:50:30PM -0500, kfogel@collab.net wrote:
> I'm trying to figure out how the VCP source driver for CVS handles
> branches and tags that touch multiple files (for example, when you
> branch or tag an entire directory tree). Just based on reading the
> documentation, I get the feeling that VCP deals with one file at a
> time, and doesn't attempt to unify CVS branch and tag events even
> though the destination repository might support that.
right, the source driver is responsible to extract `per-file-revision',
and VCP::Filter::changesets will group then by user-defined rules.
by the default the rule is: same author, comment, branchi , and within 60sec.
> The reason I'm asking is to find out how much our own cvs2svn tool
> duplicates the work of VCP.
iirc cvs2svn requires direct access to the history file and the repository?
> > I've just converted VCP from its original perforce depot to
> > http://svn.clkao.org/revml/. the trunk has the svn destination
> > driver I added.
>
> 'svn log -v' on that repository shows no copied files. There were a
> bunch added at the beginning, and then it's just "A"s and "M"s after
> that. (Are there just no branches / tags?)
>
> I guess there are just no branches in that repository, then.
right, there's no branch for the vcp development in p4. and i haven't
cp the tags in svn dest driver yet.
I just fixed a but in the cvs source driver preventing cvs repositories
that are created by `cvs import' initially to be converted. right now
medium size repository (with 500 revisions) and a few branches are tested.
i still need to test larger stuff with more branches.
> > as for performance, it took one hour to convert 612 revisions
> > from p4 to svn. currently vcp invokes `svn' to do everything.
> > native perl binding to subversion could help a lot.
>
> Yes -- or, output to an svnadmin-style dumpfile, might be much faster
> to do that and then load the dumpfile. (A RevML-to-SVNDumpfile
> converter, I guess!)
cool. i just skimmed the svndumpfile format, and it seems the
nodes for change is using the full content of the file instead
of delta. is there any reason for it? also, is it possible for
loading to non-empty repository?
Cheers,
CLK
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sun Jun 8 07:01:29 2003