We need to be more serious about the way we manage our CVS repository.
Having people changing one branch or another and merging pieces here and
there when they feel like it is just plain wrong. CVS is crappy enough*
that it doesn't remember any merge you've done. When you merge from one
branch to another, it always considers the root of the branch to be the
base of the merge. But if you've already merged since then it is not.
Taking the root of the branch as the base, CVS then tries to apply all
changes any way and most likely ends up with conflicts all over the
place.
The bottom line is that we need to know in what state the branch exactly
is, which can be done with tags. A long time ago I sent out a message
explaining how to make a merge of the whole tree from one branch to
another. But being the only one to stick to that plan it doesn't work.
There are two possible usage policies:
1) only merge the whole tree every once in a while with appropriate
tagging
2) only merge pieces EVERY TIME a change is made on one branch and it
applies to the other
I had started with 1) but it hasn't worked. I'm ok with changing for 2).
But anything in between, which is what has been done so far, is NOT a
good option.
Let's pick one or the other and stick with it!
--
Arnaud Le Hors - IBM, XML Standards Strategy Group / W3C AC Rep.
*: yes, I consider CVS crappy, it does the bare minimum and the only
reason it is used so much is because it's free and "good enough" in
general. But there are many things for which it is just not good enough
to me.
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
