Running and debugging

Workspaces and git

A flow that works on code lists its repositories. Each run's agents and scripts then work in a workspace: a checkout of each repository, all on a git branch of the run's own, so runs never step on each other.

Adding repositories

1. Open the flow's settings

   Press Settings in the board's toolbar.

2. Add a repository

   Under Repositories, choose the host: GitHub takes org/repository, Other takes an https URL. Optionally give a Base branch to start from. Each listed repository has Remove.

3. Set a branch prefix (optional)

   Branch prefix names the branches runs make. It applies to runs started from then on.

Branches

Every run works on a branch of its own, named after the prefix and the run:

spoletta/sentry/run-41

• Every process of a run works on the same branch and picks up where the one before left off: the agent that fixes a bug and the one that opens the pull request see the same checkout.
• When a flow process runs once per item of a list, each item gets its own branch, so items running at once don't meet.
• A re-run gets a branch of its own, starting from where the original was when its process started. See Re-runs and follow-ups.

Commits

Whatever a step leaves uncommitted is committed for it, with the run, the step and the agent's session recorded in the commit. Each step in the run panel shows Worked on the branch, the commits it made per repository, from one commit to another, and whether it left changes uncommitted. Pushing and opening a pull request is up to the flow: an agent asked to “push the branch and open a pull request” does so.

Once a run is over, its checkouts are released, but the branches and agent sessions are kept, so a follow-up message or a re-run makes them again from the branch.

Which steps get a checkout

In a script, $WORKSPACE is the directory holding a checkout of each repository, named as the repository is, and $BRANCH is the run's branch.