Re: CVS2GIT possible parents of Tags / Branches

James Blackburn wrote:
> Using cvs2git (2.3.0 and current trunk) on an old CVS repository, it
> seems be mis-calculating tag & branch parents frequently.
> 
> For example:
> 
> 0     FP_BINUTILS_DEV_209_RELEASE    tag        . .trunk.
>       # 'FP_BINUTILS_DEV_209_RELEASE' is a tag in 14302 files, a
> branch in 0 files, a trivial import in 0 files, a pure import in 0
> files, and has commits in 0 files
>       # Possible parents:
>       #     MOVE_ONLY_BRANCH : 14188
>       #     .trunk. : 11758
> 
> It's detected FP2010_FNSC_MOVE_ONLY_BRANCH as a better parents (though
> I'm not sure how this is possible). Taking an arbitrary file as an
> example, I see:
>         FP_BINUTILS_DEV_209_RELEASE: 1.6
>         ....
>         FP_BINUTILS_DEV_191_RELEASE: 1.6
>         MOVE_ONLY_BRANCH: 1.6.0.2
> 
> Is this expected? Is there any way of working out why it thinks the
> development branch is better than trunk?

The possible parent decisions are based on statistics over the whole
repository, so giving us an example from a single file is not enough to
prove that it is broken.  (That's not to claim that it isn't *broken*;
just that it isn't *proven*!)

What might be confusing (though it is not the case in this file) is that
"siblings" of a branch are considered possible parents.  This is because
of a CVS quirk.  Suppose BRANCH1 is created from trunk, then BRANCH2 is
created from BRANCH1:

-*--*---*-------*--- trunk
     \
      --*--*-------- BRANCH1   1.2.2
            \
             ------- BRANCH2   1.2.2.2.2

In files for which there was a commit on BRANCH1 before BRANCH2 was
created, the branching tree looks like the above and is unambiguous.
But in a file where there was no commit on BRANCH1 before BRANCH2 was
committed, CVS records a branching tree like this:

-*--*---*-------*--- trunk
     \
      |-------------- BRANCH1   1.2.2
      |
       -------------- BRANCH2   1.2.4

with BRANCH1 and BRANCH2 siblings, both directly branched from trunk.
In fact, sometimes the numbering order of BRANCH1 and BRANCH2 is not
even correct.

So, please give us a small but complete test case that illustrates the
problem that you have so that we can analyze it properly.

Yours,
Michael

------------------------------------------------------
http://cvs2svn.tigris.org/ds/viewMessage.do?dsForumId=1670&dsMessageId=2650961

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