[darcs-users] Re: announcing darcs 1.0.0

Catalin Marinas catalin.marinas at arm.com
Thu Nov 11 16:46:19 UTC 2004

David Roundy <droundy at abridgegame.org> writes:
> On Wed, Nov 10, 2004 at 01:43:53PM +0000, Catalin Marinas wrote:
>> A related question - why is this tagging needed? It is the unique
>> bkrev id. Well, maybe it is needed, in case somebody wants to retrieve
>> a patch using it but I don't thinks so.
> It's needed if anyone wants to retrieve a version identical to the bkcvs
> version, since darcs doesn't necesarily maintain the order of
> patches.

I just noticed that it is completely broken. See below.

> Hmmm.  It does indeed seem that cvs2darcs is behaving oddly.  Either that
> or bkcvs is behaving oddly.  Or I guess the third possibility is that cvsps
> is behaving oddly.  cvs2darcs uses cvsps to get the changeset information,
> including whether there is a tag associated with each version, so either
> darcs is failing to parse properly the cvsps output, or it is giving wrong
> output.  I don't have time right now to check which is the case.

I tried cvsps and it doesn't generate the patchsets properly. In
bkcvs, the revision log is stored as a log for the ChangeSet file and
the rest of the files contain the (Logical revision ...) text in their
log. Cvsps doesn't properly match those. You can get several patches
for the same revision or you can get them in a different order. Even
worse, the ChangeSet log (together with the BKRev tag you use) for a
given logical revision can appear before the actual patch for the
files corresponding to that patch (that's why the bkrev tagging is

There might be a possibility to use the cvsps generated patches but
re-sorted after the logical revision and do a "cvs log ChangeSet -r
<logical_rev>" to get the actual log message.

Another possibility would be to use the patches sent to the
bk-commits-head mailing list (also accessible through
gmane.linux.kernel.commits.head) and apply them one by one. I'm not
sure if they can be cleanly apply in the order they are posted though
(I don't know bitkeeper's internals).


More information about the darcs-users mailing list