Platform Team/gatch

From Sugar Labs
Jump to: navigation, search

Summary

Support long or ever standing, all time being on top, downstream patchsets in a Git repository. This tool is intended to be lightweight and support only long or ever standing downstream patches. For the rest of use cases, look for the alternatives.

Install

Clone Gatch repository:

git clone git://git.sugarlabs.org/sdk/gatch.git <INSTALL-PATH>

Add gatch command to the PATH environment variable, e.g., in ~/.bashrc file (you need to relogin):

echo PATH=<INSTALL-PATH>:$PATH >> ~/.bashrc

Usage

This section operates with the following terms:

  • upstream remote
    The name (from git remote command output) of repository that contain upstream branch;
  • upstream branch
    Git branch in upstream remote repository that is beeing used as a basis for downstream branch;
  • downstream remote
    The name (from git remote command output) of repository that contains downstream branch, might be the same as upstream remote;
  • downstream branch
    Git branch in downstream remote that is based on upstream branch and contains downstream commits all time on top of ones from upstream branch.

Initiate downstream branch

Add upstream remote (if it is not the same as the current remote, origin) for repository downstream branch will be based on:

git remote add <UPSTREAM-REMOTE> <UPSTREAM-URL>

Create downstream branch:

git fetch <UPSTREAM-REMOTE> --no-tags
git checkout <UPSTREAM-REMOTE>/<UPSTREAM-BRANCH> -b <DOWNSTREAM-BRANCH>

Register downstream branch as a Gatch branch:

gatch init <UPSTREAM-REMOTE> <UPSTREAM-BRANCH>

This command will create watershed commit with comment [gatch-<DOWNSTREAM-BRANCH>] Initiate downstream branch. Since this time, all downstream commits will be after watershed one, even after re-basing to the recent upstream branch.

Populate downstream commits

Use downstream branch as a regular Git branch, i.e., create new commits.

Merge upstream updates

Being in downstream branch, type:

gatch merge

This command will:

  1. fetch recent upstream branch commits;
  2. rollback downstream commits to the currently local upstream HEAD;
  3. pull upstream commits on top on currently local upstream HEAD;
  4. re-apply all downstream commits on top on new upstream HEAD.

The whole process does not break Git's consistent history, i.e., while pushing new downstream branch to the server, there is no need in --force argument.

The one obvious cave is that on the 4th step, git might complain about not possible merging. These conflict need to be solved as regular Git's merge conflicts.

Tips

Upstream merge right after clone

init command keeps all information regarding upstream repository and branch in .sweets/gatch* files controlled by git. So, it is possible to clone downstream branch and run merge command, gatch will add upstream remote if it absent in local repository.

Avoid upstream merge history

Gatch does not break Git history, thus, it is possible to see downstream commits that existed before every upstream merge. It is useful to git --hard reset sometimes, but might be annoying to see all of them in git log output. To suppress showing these commits, use git log --first-parent command.

See also

  • guilt
    A series of bash scripts which add a quilt-like interface to git.
  • stgit
    Manage a stack of patches using GIT as a backend.
  • topgit
    A different patch queue manager.

Getting involved

  • Read the HACKING file to know how to contribute with code.
  • Gatch sources.