[Gnu-arch-users] Working out a branching scheme [was: tag --seal --fix]

[Juliusz Chroboczek]

TL> So, sealing is a kind of advisory lock and you can use it to model
TL> releases.

That appears to be what I'm looking for.  Thanks a lot.  (And thanks
to Dustin for the link.)

Before I switch to arch, I'm trying to work out the branching model to
use.  I'm thinking of the following.

All development happens on foo--devel--0.7.  When 0.7.0 is released,
it is tagged as foo--release--0.7.0.  Bug fixes to the 0.7 branch
continue on foo--devel--0.7, with later tags foo--release--0.7.1 and
so on.  There are never any commits on the foo--release branch.

When a new unstable branch is created, it is tagged as
foo--devel--0.8.  Development happens on this new branch, and bug
fixes to the stable branch can go to foo--devel--0.7.

I guess what I'm trying to achieve is something equivalent to cvs
symbolic tags, which I'm used to.  If you see better ways of doing
that, I'm all ears.

Oh, by the way: there are two things that are easy with cvs that I
couldn't work out how to do with arch without checking out at least
one old version.  One is

  cvs diff -r FOO_0_7_0 -r FOO_0_7_1

which allows me to get a patch between two old versions.  The other is
to generate a patch in which all file deletions and moves are made
explicit with suppressions and additions.  In short, I'm looking for
an efficient way to do:

  tla get foo--release--0.7.0 ,foo-0.7.0
  tla get foo--release--0.7.1 ,foo-0.7.1
  diff -urN -x '*/{arch}/*' ,foo-0.7.0 ,foo-0.7.1
  rm -rf ,foo-0.7.0 ,foo-0.7.1

(And of course the ability to do cvs export, but that's already in the
wiki.)

Thanks for your help,

                                        Juliusz