Commiting Patches with CVS
Motivation
Have you ever wanted to commit only some of the changes that you've
made to few files, but not, for example, the debug code that you added
to get it to work? Still, you'd like to keep that debug code in your
working directory, you just don't want to check it in.
Do you like to check in only independent and related changes in each
commit even though you may be working on more than one thing at once?
Have you wanted to makes a quick change to a file before checking it
in, but still have the old version in your working directory?
If so, then cvs-commit-patch and its emacs interface
cvs-commit-patch-buffer can help make this process as simple and easy
as editing a diff buffer in emacs.
This code has been working reliably for me for over two years.
The program: cvs-commit-patch
cvs-commit-patch is an independent shell script that commits a patch
(as generated by diff(1)) to a cvs repository. It
- handles file additions/deletions,
- verifies that the patch applies cleanly even
when the working directory is un-clean,
- can be reliably interrupted (ctrl-c),
- patches and updates working directories with distinct
or overlapping changes.
It requires bash(1) and a few utilities from the patchutils
distribution.
The emacs Interface: cvs-commit-patch-buffer.el
cvs-commit-patch-buffer.el is an emacs interface to cvs-commit-patch.
It allows you to just hit C-c C-c in any patch buffer to apply and
commit only the changes indicated by the patch, regardless of the
changes in your working directory.
One method of working with cvs-commit-patch-buffer is to just M-x
vc-diff a file then kill, split or edit the resulting hunks and to
then hit C-c C-c to commit the patch. The other is to use PCL cvs
mode to tag many files and then to diff them into a buffer which can
again be edited and committed.