threads / discuss / 3374

git-cvs-import and branches

Subject: git-cvs-import and branches

## tl;dr

2 messages between Feb 17, 2006 and Feb 17, 2006.

replies: 1people: 2as markdown or json

Martin Mares· Feb 17, 2006, 19:55 UTC · lore

Hello!

I'm git-importing a rather large CVS repository with lots of branches and although the mainline is imported correctly, the branches aren't. They usually contain some changes introduced to the ancestor branch after the point the new branch has been tagged.

I haven't studied git-cvsimport and cvsps in much detail yet, but it seems that git-cvsimport forks off the branch at the first patchset reported by cvsps which mentions the branch and copies the current state of the ancestor branch at that time.

This doesn't seem to be correct, since many changes might have happened to the ancestor branch since the real branching point. However, since cvsps doesn't report the branching points (and they aren't exact points anyway since tagging is not atomic in CVS), I don't see how to make cvsimport find the right starting revision.

Does anybody have an idea how to solve this?

				Have a nice fortnight

-- 
Martin `MJ' Mares   <mj@ucw.cz>   http://atrey.karlin.mff.cuni.cz/~mj/
Faculty of Math and Physics, Charles University, Prague, Czech Rep., Earth
COBOL -- Completely Outdated, Badly Overused Language

David Mansfield· Feb 17, 2006, 20:35 UTC · re: Martin Mares · lore

Re: git-cvs-import and branches

On Fri, 2006-02-17 at 20:55 +0100, Martin Mares wrote:

Show 17 quoted lines

> Hello!
> 
> I'm git-importing a rather large CVS repository with lots of branches and
> although the mainline is imported correctly, the branches aren't. They
> usually contain some changes introduced to the ancestor branch after
> the point the new branch has been tagged.
> 
> I haven't studied git-cvsimport and cvsps in much detail yet, but it seems
> that git-cvsimport forks off the branch at the first patchset reported by
> cvsps which mentions the branch and copies the current state of the ancestor
> branch at that time.
> 
> This doesn't seem to be correct, since many changes might have happened
> to the ancestor branch since the real branching point. However, since
> cvsps doesn't report the branching points (and they aren't exact points anyway
> since tagging is not atomic in CVS), I don't see how to make cvsimport
> find the right starting revision.

Ugh. I understand the problem but really don't have any solutions off-hand. Unfortunately, I dying under a 'real work' load at the moment, so I have to apologize and say, 'good luck.'

David