[cvsnt] Re: virtual branches ?

[Matt Schuckmann]

Tony Hoyle wrote:
> Matt Schuckmann wrote:
> 
>> Surely there is a need for it or else people like me and whomever 
>> implemented the Move Branch to Head macro in WinCVS wouldn't be doing it.
>>
>> If I remember right if a file changes on the main line but not on a 
>> branch when you do a merge from the mainline to the branch a new 
>> revision is created on the branch that is identical to the revision on 
>> the mainline. In my mind this is wasteful and unnecessary. Plus it makes 
> 
> 
> Revisions are a few bytes.. even if it created hundreds, even thousands 
> of them it wouldn't make any difference.
> 
> What is important is reproducability - you need to be able to checkout a 
> branch and get *exactly* what you got before, otherwise any testing done 
> is meaningless.  That's the whole point of a source control system - and 
> is why it's not uncommon to have the whole build system under revision 
> control, not just the source code.

No I'd say that's what a tag is for. A branch is an alternate line of 
development for one or more files. Anybody can check in new revisions 
one or more of the files with a branch tag and all of the sudden you get 
something different when you check out a branch. What I'm suggesting is 
  really no different checking in new revisions on a branch or 
performing a merge from trunk to branch.

> 
> You can't simply move the branches around in many cases - the branch is 
> a delta from the revision that it was created at.  Moving branches with 
> revisions on is error prone and IMO dangerous.  Magic branches handle 
> the non-revision case for the (rare) cases that it's useful (the 
> original example was template configuration files with customer specific 
> changes to a few of them).
> 

I don't remember asking to move branches with revisions on them, that 
would be a stupid thing to do. I'll try to say it again: It would be 
nice to have a command that floats branch tags for files which do not 
have revisions on that branch yet. This is very similar to the magic 
branch  concept with the exception that you have control over when the 
float happens. This would be an operation done just before merging from 
the trunk to a branch.
The idea being you are starting work on some task so you create a 
branch, you  modify a few files and check them in to your branch. Mean 
while changes have continued to happen on the trunk. At some point you 
feel that the changes on the trunk are stable enough that you want to go 
through the "expense" and additional testing of bring those changes into 
your branch for your benefit and testing. First you float the branch tag 
for the files which have changed on the trunk but not in your branch 
then you preform the normal merge operation, maybe it should just be 
part of the merge with a special option to indicate that you want to 
float unbranched files.

Matt S.

> Merging is a nontrivial operation for a reason - you can't just take two 
> random branches and stick them together without operator intervention.
> 

Where did that come from, it has no bearing on anything that has been 
said so far.

> Tony