librelist archives

« back to archive

Error revealed by vendor.testrepo?

Error revealed by vendor.testrepo?

From:
Eric S. Raymond
Date:
2014-12-26 @ 17:57
Our conversion of vendor.testrepo passes a cvsconvert check against its CVS.

What bug in the vendor-branch behavior do you think it's exposing?
-- 
		<a href="http://www.catb.org/~esr/">Eric S. Raymond</a>

Re: [cvsfastexport] Error revealed by vendor.testrepo?

From:
Loz
Date:
2014-12-26 @ 18:27
Subset tag will probably paper over the problem. Imagine the repo didn't 
have the "lost-repo-state" tag. There wouldn't be a commit for this state 
in the git repo.


> On 26 Dec 2014, at 17:57, Eric S. Raymond <esr@thyrsus.com> wrote:
> 
> Our conversion of vendor.testrepo passes a cvsconvert check against its CVS.
> 
> What bug in the vendor-branch behavior do you think it's exposing?
> -- 
>        <a href="http://www.catb.org/~esr/">Eric S. Raymond</a>

Re: [cvsfastexport] Error revealed by vendor.testrepo?

From:
Robert de Bath
Date:
2014-12-27 @ 00:33
Oh, yes; we're getting some 'papering' happening here.

Most of the tags are looking right in the CVS CVS, but do a "git lola" and
you can see that the problems are being "fixed" by adding a last commit to
bend it into shape.

Still, it does look like it's on the right path, possibly by propagating 
the same rules back through the history until this branch joins onto an 
earlier one.  That would push the 'fixer' commit to the junction between 
the branches which IMO would better match the history that you can extract
from the CVS repo (ie start at a tag and go back in time from there).

As I'm putting up notes here I should also probably point out that there's
something funny happening with the ".cvsignore"/".gitignore" files in the 
emacs CVS repo; cvsconvert won't be seeing it 'cause it ignores that file.
Also, in CVS CVS, there's the CVS, ELIB, ZLIB and GNU_AUTOMAKE tags, 
they're missing too !? And one last nasty I've come across: It seems that 
it's acceptable to use the "rcs -oXYZ" command on files sometimes 
(normally with the "cvs admin -o" program); that deletes old deltas in the
RCS file ... so there might not be a version 1.1 in the file, or versions 
may be skipped. I'm not sure yet if that last one will cause problems ... 
but it's not a nice thing to find!


NB: git lola --> git log --graph --decorate --pretty=oneline 
--abbrev-commit --all --date-order


Robert de Bath
.

________________________________________
From: cvsfastexport@librelist.com [cvsfastexport@librelist.com] on behalf 
of Loz [loz@flower.powernet.co.uk]
Sent: 26 December 2014 18:27
To: cvsfastexport@librelist.com
Subject: Re: [cvsfastexport] Error revealed by vendor.testrepo?

Subset tag will probably paper over the problem. Imagine the repo didn't 
have the "lost-repo-state" tag. There wouldn't be a commit for this state 
in the git repo.


> On 26 Dec 2014, at 17:57, Eric S. Raymond <esr@thyrsus.com> wrote:
>
> Our conversion of vendor.testrepo passes a cvsconvert check against its CVS.
>
> What bug in the vendor-branch behavior do you think it's exposing?
> --
>        <a href="http://www.catb.org/~esr/">Eric S. Raymond</a>