Re: squidng branch management

From: Henrik Nordstrom <hno@dont-contact.us>
Date: Tue, 26 Dec 2000 16:48:40 +0100

As you probably know the SquidNG branches was brought up to date wrt
HEAD as part of the move to squid.sourceforge.net.

The questions about how to best manage the SquidNG branches still
remains, and as usual the problem is twofold.

1) Development on separate branches makes it a whole lot easier to merge
stuff back into Squid-HEAD.

2) Long-term development is a bit troublesome if done on separate
branches, and might not bring all the wanted synergies between different
but related (or even unrelated) developments.

There are two ways to manage complex development threads involving more
than one development without a clear-cut dependency hierarchy.

a) Create a composite branch, and continue the development there, and
then manually split it again when it is time to merge stuff back into
HEAD.

b) Continue development on separate branches, and create a new branch
that inherits changes from two or more other development branches but
which does not receive any direct changes.

(a) is strongly recommended when you find that there are strong
dependencies or synergies between the branches, especially if it is a
long time before stuff is to get merged back.

(b) is recommended if the developments are mostly separate from each
other, but needs to be tested together. It does however require quite a
bit of disipline to never commit any changes directly on the composite
branch.

If both developments are active then the choice is a bit harder, and I
have no good answer to give there.

For the SquidNG project, reiserfs_raw should have been merged into
squidng-HEAD a long time ago. It is the base for most (all?) development
threads there. Now, as cbdata should be close to the point of merging
into Squid-HEAD the issue should be quite moot for reiserfs_raw as
cbdata is the only thread between Squid-HEAD and reiserfs_raw, but the
problem still applies to the later threads.

One thing which is important for making any of the approaches work is
that changes are originally committed on the branch where they belong.
If you have branched off from another development thread and find a
problem in that code, fix it there. Doing it and leaving it in another
sub-branch is not a good idea.

/Henrik

Adrian Chadd wrote:
>
> On Wed, Nov 08, 2000, Yury Shevchuk wrote:
>
> > > When is this going to make it into the squidng HEAD branch?
> >
> > In fact, I don't understand what other developers need at this point.
> > I can merge the squidng reiserfs_raw brinch to squidng HEAD. Or I can
> > move it to reiserfs_raw branch of squid.sourceforge.net. Please
> > advice what is better?
>
> Hrm. I'll talk to Henrik about it when he gets back from holiday. I'd
> like to bring the squidng devel branch uptodate and then get the
> squidng HEAD code uptodate with the latest squid code so I can start
> to bring it across to squid-HEAD ..
>
> grr, not enough time ... :-)
Received on Tue Dec 26 2000 - 08:49:32 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:13:06 MST