Re: vendor branching and merge

[Eric Siegerman]

On Mon, Feb 25, 2002 at 09:06:36PM -0500, Greg A. Woods wrote:
> [ On Monday, February 25, 2002 at 12:01:26 (-0500), Eric Siegerman wrote: ]
> > If you "add+commit" a new file to the vendor branch, cvs doesn't
> > set the RCS default branch to 1.1.1, as creating the file via
> > "import" would do.
> 
> Ah, yes, that's definitely one of the issues.....
> 
> >  NOT a bug, IMO; but it is a difference that
> > should be borne in mind.
> 
> Hmmmm....  it might be a bug -- it depends I guess on what it means to
> "work" on the vendor branch.

It would be nice to have CVS deal with this automatically.  But
how is it to know when the sandbox is on a vendor branch?  The
usual 1.1.1 case is easy, but in general, it's impossible, given
the existence of import's "-b" option.

More fields for val-tags?  Turn it into a proper (text) database
of info about tags, including "this tag represents a vendor
branch".

Hey, this could be cool.  Another per-(branch-)tag data item
that'd be useful is a flag that says, "this branch is dead".  So
when you want to merge and terminate a branch, you could make
that explicit, instead of the current "merge and forget -- and
hope everyone else forgets too".  "cvs terminate-branch" (yucky
name!) would:
  - check in "dead" revisions on the branch, à la "cvs rm"
  - set the dead-branch flag (or, more intuitively, clear the
    active-branch flag) in val-tags

There are compatibility issues, of course, but with a bit of
thought, it might be possible to handle them cleanly.

--

|  | /\
|-_|/  >   Eric Siegerman, Toronto, Ont.        address@hidden
|  |  /
One ring to rule the mall.
        - Movie review headline, eye Magazine