New Skara feature: dependent pull requests

[Erik Helin]

Hi all,

me and Robin are looking to deploy a new feature for the jdk [0] 
repository this week: dependent pull requests. Dependent pull requests 
are used when you have a change that depends on work in a pull request 
that is not yet integrated.

For an example, lets say that you are fixing a bug and realize that you 
first need to refactor a piece of code in order to make the bugfix 
cleaner. To aid reviewers you do *not* want to create a pull request 
containing both the refactoring and the bugfix - they are distinct 
changes and should be reviewed independently. You create a pull request 
for the first change, the refactoring, but then what? You cannot create 
a pull request for the bugfix only since the refactoring is not yet 
integrated. The review of the refactoring might take a while and during 
this time you cannot receive reviews for the bugfix.

Dependent pull requests will help in the above scenario. When a pull 
request is created then the Skara bots will automatically create a 
branch in the upstream repository named `pr/<ID>` where `<ID>` is the 
id of the pull request (e.g. `17`). This `pr/<ID>` branch can then be 
used as the target branch when creating a pull request. A pull request 
with a dependency is automatically targeted to the dependency's target 
branch when the dependency is integrated. For example, if the first pull 
request (with id `17`) targets the `master` branch and the second pull 
request targets the `pr/17` branch, then the second pull request will be 
automatically re-targeted to the `master` branch when the first pull 
request (with id `17`) is integrated. It is not possible to integrate a 
pull request with a dependency on an open pull request (i.e. the 
dependency must be integrated first). The branches named `pr/<ID>` are 
automatically removed when the corresponding pull request is closed.

Dependent pull requests have been enabled for a long time for the skara 
repository [1] and an example of a pull request with a dependency can be 
seen here [2].

For those of you using the Skara CLI tools [3] then `git-sync` will by 
default *not* sync branches named `pr/<ID>` to the personal fork. If it 
did then you would have to delete all `pr/<ID>` branches in your 
personal fork yourself. This default behavior can be overridden by 
passing `--branches` to `git-sync`.

Let us know on the skara-dev [4] mailing list if you have any questions!

Thanks,
Erik and Robin

[0]: https://github.com/openjdk/jdk
[1]: https://github.com/openjdk/skara
[2]: https://github.com/openjdk/skara/pull/1087
[3]: https://wiki.openjdk.java.net/display/SKARA/CLI+Tools
[4]: https://mail.openjdk.java.net/mailman/listinfo/skara-dev