Permalink
Fetching contributors…
Cannot retrieve contributors at this time
129 lines (83 sloc) 4.24 KB
Conventions? Are you kidding? OK fine.
Code Branching Model
====================
The master branch is what we consider the main-line of development,
and the last, non-rc tag on master is the most recent stable release.
Any branch with a "tmp/" prefix might be rebased (often), so keep that
in mind when using or depending on one.
Any branch with a "tmp/review/" prefix corresponds to a patchset
submitted to the mailing list. We try to maintain these branches to
make the review process easier for those not as familiar with patches
via email.
Current Trajectory
==================
Now that we've finished the 0.29.1 release, we're working on 0.30, and
although we're not certain which new features will be included, here
are likely candidates:
- Support for transferring saves between repositories and rewriting
branches.
and these are also under consideration:
- Better VFS performance for large repositories (i.e. fuse, ls,
web...).
- Incremental indexing via inotify.
- Smarter (and quieter) handling of cross-filesystem metadata.
- Support for more general purpose push/pull of branches, saves, and
tags between repositories. (See the bup-get patch series.)
If you have the time and inclination, please help review patches
posted to the list, or post your own. (See "ways to help" below.)
More specific ways to help
==========================
Testing -- yes please.
With respect to patches, bup development is handled via the mailing
list, and all patches should be sent to the list for review (see
"Submitting Patches" below).
In most cases, we try to wait until we have at least one or two
"Reviewed-by:" replies to a patch posted to the list before
incorporating it into master, so reviews are an important way to help.
We also love a good "Tested-by:" -- the more the merrier.
Testing
=======
You can run the test suite much more quickly via "make -j test" (as
compared to "make test"), at the expense of slightly more confusing
output (interleaved parallel test output), and inaccurate intermediate
success/failure counts, but the final counts displayed should be
correct.
Individual non-Python tests can be run via "./wvtest run t/TEST" and
if you'd like to see all of the test output, you can omit the wvtest
run wrapper: "t/TEST"
Individual Python tests can be run via "./wvtest run ./wvtest.py
lib/bup/t/TEST", and as above, you can see all the output by omitting
the wvtest run wrapper like this: "./wvtest.py lib/bup/t/TEST"
Submitting patches
==================
As mentioned, all patches should be posted to the mailing list for
review, and must be "signed off" by the author before official
inclusion (see ./SIGNED-OFF-BY). You can create a "signed off" set of
patches in ./patches, ready for submission to the list, like this:
git format-patch -s -o patches origin/master
which will include all of the patches since origin/master on your
current branch. Then you can send them to the list like this:
git send-email --to bup-list@googlegroups.com --compose patches/*
The use of --compose will cause git to ask you to edit a cover letter
that will be sent as the first message.
It's also possible to handle everything in one step:
git send-email -s --to bup-list@googlegroups.com --compose origin/master
and you can add --annotate if you'd like to review or edit each patch
before it's sent.
For single patches, this might be easier:
git send-email -s --to bup-list@googlegroups.com --annotate -n1 HEAD
which will send the top patch on the current branch, and will stop to
allow you to add comments. You can add comments to the section with
the diffstat without affecting the commit message.
Of course, unless your machine is set up to handle outgoing mail
locally, you may need to configure git to be able to send mail. See
git-send-email(1) for further details.
Oh, and we do have a ./CODING-STYLE, hobgoblins and all, though don't
let that scare you off. We're not all that fierce.
Even More Generally
===================
It's not like we have a lot of hard and fast rules, but some of the
ideas here aren't altogether terrible:
http://www.kernel.org/doc/Documentation/SubmittingPatches
In particular, we've been paying at least some attention to the bits
regarding Acked-by:, Reported-by:, Tested-by: and Reviewed-by:.