Re: git-branch, older repos and more confusion

On Thu, Apr 12, 2007 at 05:05:24PM +1200, Martin Langhoff wrote:

Which means I'll have to write a little list that goes "if your git is
version X, do this, if it's version Y do that". :-(

Yes, that is frustrating. If you are just working with master/origin, I
think you can get away with it due to the default configuration created,
but otherwise I'm not sure it's avoidable. I seem to recall a thread
about this a few months ago (between me and Bill Lear, maybe?), but I
don't recall that there was any especially good conclusion.

For example, creating a new "custom" branch on the repo used to be
trivial and left us with a setup that was simple and safe. For
example, if you wanted to work on a custom v1.9:

- cg-clone <repo>#v1.9-maint mydir
- cd mydir
- cg-branch-chg origin <repo>#1.9-clientname
- cg-push     ## this creates the new head on the repo
- cg-branch-add v1.9-maint <repo>#v1.9-maint mydir

and from there onwards cg-update / cg-push worked on the custom
branch, and cg-update v1.9-maint would merge from the maint branch.

Any hints as to how to run such workflow on v1.5.x?

This will seem like repetition, but it's like this:

  git-clone <repo> mydir
  cd mydir
  git-checkout --track -b v1.9-clientname origin/v1.9-maint
  git-push origin v1.9-clientname    # to create the new head remotely

When you do a git-pull without arguments from the v1.9-clientname
branch, it will first do a fetch of origin (grabbing the entire state,
including the v1.9-maint branch) and then merge in origin's v1.9-maint.

When you do a push without arguments, it will push every head you have
in common with origin. Because we pushed the v1.9-clientname branch
once, it was created and will be part of subsequent pushes.

-Peff