This file describes how to contribute code to Kresus. Note that many other contributions are valuable and would be glady accepted in all of the following domains: design, user interface, user experience, translations, documentation, tutorials. If you want to get started with those, please head to our community forum. The only limit is the imagination!
These rules and recommendations can change over time, and should change. If you have any ideas for improving them, please file an issue or open a merge request!
- Branch off
master
. - One feature per commit.
- In case of requests for changes, amend your commit.
- First, install the app's dependencies:
npm install
- Copy
config.example.ini
toconfig.ini
and set values for your local development environment. - Start development mode:
make dev
. This will automatically build the server and client files, spawn the main server on localhost:9876, (and reload it whenever a server source file is changed), spawn a client server on localhost:8080 and opens the index page on a browser (which gets reloaded every time a client file is touched).
Alternatively, you can use make watch
which will just automatically recompile
the files without auto-spawning servers.
If watching doesn't work, under Unix based operating systems (Linux, MacOS), you might need to increase the number of inotify nodes.
A series of tests are shipped with the code to avoid regressions. They are
located in the ./tests
folder.
Some tests require a valid install of Weboob to work properly, some others do not.
For that, the weboob related tests are disabled if the environment variable
KRESUS_WEBOOB_DIR
is not set. Some other tests are disabled if this
environment variable is set.
To ensure all the tests pass, you need to run the test command twice, once with
KRESUS_WEBOOB_DIR
set, once without. For example:
npm run check:test
KRESUS_WEBOOB_DIR=/path/to/weboob npm run check:test
We use the package.json
file in a reproducible way, specifying the exact
version to use. Please make sure all version numbers are exact in
package.json
, thus using no version ranges specifiers like ~
, >
etc.
To not have shell scripts in package.json
, we use scripty
: every command
that has form a:b:c
in the package descriptor file and that's sent to
scripty
will run the script scripts/a/b/c.sh
or scripts/a/b/c/index.sh
automatically.
master
contains all changes in the current development version, including some experimental features that could break in production.builds
contains a stable version of the code for releases.
- Please note that not every feature can make it into Kresus. New features add complexity and usually the maintenance burden is carried by the core team. Also, Kresus tries to limit its scope to doing one thing well.
- If you're thinking about a new feature, see if there's already an issue open about it (but don't sweat it too much if you can't find one!), or open one otherwise. This will ensure that everybody is on track for the feature and willing to have it in Kresus.
- One commit per feature.
- Branch off the
master
branch. - Rebase when you're close to landing, to make sure there's no merge skew risk. We don't do merge commits, because they break bisection and add a lot of noise in the commit history.
- Test your code with
make check
. This also runs linting and a few consistency checks. - For client changes, it is highly recommended to run extensive tests, and to note if there are tests known to fail. In particular, when touching a form, it's strongly advised to try to use invalid data, run it, close it, enter data and close it and re-open it, etc.
- We'll look at your MR and might ask for a few changes. In this case, use your best judgement to consider wheter it's smarter to create new commits or to amend existing ones. When the final result looks good, we may ask you to squash the WIP commits into a single one, to maintain the invariant of "one feature, one commit".
- Core contributors:
nicofrand
,ZeHiro
,Phyks
. Core contributors can review and merge MRs, and have full power on the repository, including but not limited to push access on themaster
branch. - Module owner:
bnjbvr
(if alive). They get the last word and can veto the progression of a particular merge request, which should only happen in last resort if no cooperative solutions have been found otherwise.
- More than one person must have the commit accesses on
master
. - All code changes must pass through the process of review. As a matter of
fact, it is not allowed to push directly on the
master
branch. All changes must go through a merge request. - All merge requests must be reviewed and approved by at least one core
contributor before they can be considered for a merge. The marking of a merge
request with the
ship-it
label indicates that a merge request can be merged by the author, once the remaining issues / remarks have been addressed; of course, if other questions arise, the author can ask for another round of review (and unmark the MR asship-it
). - Merge requests shouldn't be merged no less than one day after they've been proposed, to make sure people have time to test them and think about all the possible implications they could have.
- Bug fixes can be merged without too much wait.
- New features or anything involving architecture discussions and design interactions should be reviewed and approved by the module owner.
- If a review takes more than a few days, it is appropriate to gently ping the reviewer(s). Note that they might be busy in meat space. Not getting a review isn't a reason to merge any time sooner.
- Failing to respect these rules may result in losing the right to merge, after a first warning strike and a discussion between contributors.
- Update the version number in the package.json file on the
master
branch. - Checkout the
builds
branch and merge frommaster
withgit checkout builds && git merge -X theirs master
(which will always take master changes). - Run
make release
. - Check
git status
, unstage unwanted changes, and commit withBuild;
in the commit message. - Run
git tag 0.14.0
with the version number. - Push the
builds
branch and the tag.
- Just after this on the same branch, run
npm publish
. - Test npm release with
npm install --prefix /tmp kresus
and run Kresus from there.
- Run
make docker-release
(ensure it doesn't use cached images). docker tag bnjbvr/kresus:latest bnjbvr/kresus:0.14.0
with the right version number.docker push bnjbvr/kresus
- Write a blog post for the release:
- check commits.
- don't talk too much about implementation details unless a lot of work has been done in a particular area, e.g. tests, migrating DB, etc.
- give visibility to non-technical contributions too.
- format the blog post so all images etc. are served locally
- add Pelican metadata.
- Update the demo on demo.kresus.org with the docker image, make sure it still works.
- Create social media messages for Mastodon / Twitter and publish them with a link to the blog post.
- Ideally, re-publish social media updates a few hours / days later.
- Let package maintainers know about the update, and try to give instructions to make it easier to ugprade their packages (ArchLinux / YNH).