Why to have an alpha branch?

[Brian Rice]

On Apr 1, 2005, at 4:37 PM, Paul Dufresne wrote:

> I hope you don't mind too much I put your message on the list without 
> asking you first, but that's a very good occasion to tell everybody, 
> that I did not receive any answer on my previous email to the list 
> asking that we stop using a completely separated alpha branch, but 
> rather that we begin to tag much, much more often.
> I have the feeling that part of the reluctance to this idea is that 
> maybe (I'm not sure it is the case) CVS completely duplicate the 
> repository for each tag. My understanding of darcs is that it just add 
> some kind of empty patch (but with full context) for each tag. Which 
> seems to take very few place.

No, there's an actual darcs tag command, but I haven't used it, mostly 
because I haven't had time or pressure to do it. I do not think it 
would help other users, though, as they would have to discover the 
tags. I'd rather just say "here's the stable repository" and have them 
update from it.

> So basically what I was suggesting was to use a version number (maybe
> just one number). When main seems stable, we would tag it with the next
> number, but prefixed with an 'A' for Alpha. And, possibly, milestone 
> versions
> with an 'M'.

I think I'd rather just number and mark stable-or-not.

> I still suggest to put generated files (VM and images), in a separete 
> repository, that would use the same versions numbers as the other, but 
> would contain only files for 'A'lpha tags and 'M'ilestones.

That would be very difficult for me, to have to coordinate two separate 
repositories like that. I won't do it.

> Now on the current situation, I begin to realize that you apply all
> patches to main, then alpha (later if they can be problematic). I was 
> thinking that there was some (semi-automatic) script going on, that 
> was just erasing alpha and replacing the files by main. But your reply 
> make me suspect it is not
> the case.

I selectively pull patches from main to alpha, manually. There is no 
erasure or copying. It's all just darcs operations. That's why it works 
so well for me, is that I just pull what works and leave what has not 
been tested alone (ideally - I still make mistakes).

> That said, I did stop very recently to test my patches against alpha
> recently. I think because I would had to give a different patch for 
> alpha than for main. Probably what make me suggest to have just one 
> repository, but use tags. But I will get alpha back, and test my patch 
> against it again. But this
> is not the long term way I hope to have.

I don't understand this. Actually, most of what you say is pretty hard 
to understand, which is making this very difficult to reply to.

If it takes a page to explain, it's more complicated, which means more 
work for me, and I have plenty of that.

--
Brian T. Rice
LOGOS Research and Development
http://tunes.org/~water/