On 09/13/2012 12:30 AM, Nico Kadel-Garcia wrote:
> I'm helping upgrade a 12 year old CVS repository to use modern source
> control, and have run into a bit of a surprise. Running either cvs2svn
> or cvs2git, I'm finding a couple of old CVS files that were hand
> edited, 6 hears, and cause cvs2svn to fail with this kind of error.
>
>
> /data/cvs/tusk/ocw/tmpl/Attic/header,v
> Processed 5345 files
> Pass 1 complete.
> ===========================================================================
> Error summary:
> ERROR: '/var/www/cvs/module/badfile,v' is not a valid ,v file
>
> This is with cvs2svn 2.3.0, Subversion 1.6.12 on Ubuntu or Subversion
> 1.6.11 on RHEL. I've also tested with Subvesion 1.7.4 on RHEL, same
> problem. My problem is that the CVS files check out OK with plain CVS,
> and my clients really want to keep the history intact in their
> migration.
>
> Has anyone had fun and experience with CVS files that work in CVS, but
> break cvs2svn? Are there any particular pointers? Any guidelines on
> repairing such mangled files?
Often CVS can check recent revisions out of a repository file even if it
is corrupt, because in that case CVS doesn't need to read the whole
file. But it is likely that CVS would also have problems if you tried
to check out older revisions. For example, for a relatively thorough
test try using CVS to check out revision 1.1 and the tips of all
branches for the affected files.
It is probably possible for somebody who knows the RCS/CVS file format
to fix these files by hand and salvage some or all of the file's
history. If you need assistance, please contact me off-list.
Michael
(the cvs2svn maintainer)
--
Michael Haggerty
mhagger_at_alum.mit.edu
Received on 2012-09-13 09:53:08 CEST