Re: Dealing with repository corruption in one cvs file

Matt Ackeret wrote:
> Hi, I'm trying to convert a repository to subversion, and have one corrupted
> file from the CVS repository.
> 
> Before a few days ago, I knew *nothing* about rcs file format, and I still
> don't know much about them.
> 
> The original error I ran into was this (just removing the exact filename,
> even though I realize it doesn't really give away any info):
> 
> ERROR: './MYFILENAME.c,v' has no deltatext section
> for revision 1.8
> ERROR: './MYFILENAME.c,v' has no deltatext section
> for revision 1.7
> ERROR: './MYFILENAME.c,v' has no deltatext section
> for revision 1.6
> ERROR: './MYFILENAME.c,v' has no deltatext section
> for revision 1.5
> ERROR: './MYFILENAME.c,v' has no deltatext section
> for revision 1.4
> ERROR: './MYFILENAME.c,v' has no deltatext section
> for revision 1.3
> ERROR: './MYFILENAME.c,v' has no deltatext section
> for revision 1.2
> ERROR: './MYFILENAME.c,v' has no deltatext section
> for revision 1.1
> 
> My *original* attempt at fixing it was adding this to the end of the file
> ---
> 1.8
> log
> @DUMMY FAKE ENTRY
> @
> text
> @#error DUMMY FAKE ENTRY DUE TO REPOSITORY CORRUPTION
> @
> ---
> and so on through 1.1
> 
> That satisfied rlog, IIRC, but then gave another error during checkout.
> 
> With some help from someone else, the latest attempt was to bsaically remove
> the #error part and the additions were like this
> 1.8
> log
> @DUMMY FAKE ENTRY
> @
> text
> @@
> 
> He was able to check out manually, but for me, the co still failed:
> 
> ERROR: The command 'co -q -x,v -p1.1 ./MYFILENAME.c,v' failed with exit status=1
> and the following output:
> co: ./MYFILENAME.c,v:7948: edit script ends prematurely
> co aborted

Not very useful unless you tell us what was on and around that line number.

> So I guess I'm asking for some more ideas about how to add fake entries to 
> the end so I can do a full conversion (with *known* corruption for the very
> early versions of that file).
> 
> Or else, how can I ignore just those versions of the file in the upgrade,
> but keep all of the rest of the tags/branches?

RCS 'text' sections work like this:

The text section for the head revision (as indicated on the first line)
is the literal file contents.

The text section for every other revision on trunk is a script listing
additions and deletions of lines, to turn the successor revision back
into that revision - i.e. the script stored under 1.8 turns the full
text of revision 1.9 back into 1.8.

The text section for every revision on a branch is a script too, but
these are applied to the ancestor revision - i.e. the script stored
under 1.1.2.1 is applied to 1.1 to generate 1.1.2.1.

Max.

------------------------------------------------------
http://cvs2svn.tigris.org/ds/viewMessage.do?dsForumId=1670&dsMessageId=2423135

To unsubscribe from this discussion, e-mail: [[email protected]].

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)

iEYEARECAAYFAksJzXUACgkQfFNSmcDyxYApowCgyXSeiUeRaehJfL5Qy412oplx
WAoAoJ9FifRgg1F48guAEFKk3xNBxHpe
=J0mr
-----END PGP SIGNATURE-----