Session Framework Contributor Guide

This document outlines workflow the tips and tricks from Session Framework contributors.

Getting Started

Follow's Getting Started and Developer Workflow instructions to get your development environment set up.

Source Code Layout

The layout of the //src/session directory follows the Fuchsia Source Code Layout.

The session_manager code lives in //src/session/bin/session_manager. High level descriptions of the contents in the session subdirectories are as follows:

  • examples: example session implementations, where each example demonstrates a set of related functionality (e.g., a graphical session, a session which instantiates an element proposer)
  • fidl: internal FIDL definitions
  • tools: tools which interact with the session_manager and running session
  • lib: libraries which support the development of sessions
  • tests: integration tests for tools and bin

Tips & Tricks

fx set

Run the following command to build all libraries, binaries, and tests:

fx set core.x64 --with-base=//src/session,//src/session/bin/session_manager:session_manager.config --with //src/session:tests

Note: use --with for the tests, otherwise each run-test invocation will trigger an OTA.

fx test

Run the following command to build and execute the tests for a given area:

fx test <test>

To find the name to substitute for <test>:

  1. Find the relevant file.
  2. Copy the name of the desired unittest_package rule.

Rust IDE Integration

See the Rust Editor Configuration page for general Rust IDE setup instructions.

To provide functionality like go-to-definition and autocomplete most IDE integrations require a Cargo.toml for your project. The session framework codebase contains many projects, and it's tedious to generate each Cargo.toml file manually.

To generate Cargo.toml files for all the session framework projects, run:

# Find and build all the Rust targets.
$ fx build $(ag -G 'BUILD\.gn' 'rustc_(library|binary)\(' src/session/ | sed 's/\(.*\)\/BUILD\.gn:[0-9]\+:rustc_\(library\|binary\)("\([a-z_]\+\)") {/\1:\3/g')

# Find all the Rust targets and gen-cargo for each one.
$ for TARGET in $(ag -G 'BUILD\.gn' 'rustc_(library|binary)\(' src/session/ | sed 's/\(.*\)\/BUILD\.gn:[0-9]\+:rustc_\(library\|binary\)("\([a-z_]\+\)") {/\1:\3/g'); do fx gen-cargo $TARGET; done

Chaining Changes

Each git commit is uploaded as a separate change to Gerrit. git rebase makes it easier to split changes up into separate commits and still edit intermediate commits.

Consider the following scenario:

  1. A developer creates a feature branch and makes commits (e.g., 3 changes).
  2. The developer uploads the commits on the branch for review.
  3. The reviewers leave comments on the first two changes.

The author can then use git rebase -i HEAD~3 to select the commits they want to edit. The author can then, for each commit:

  1. Edit the code.
  2. Make sure the code still builds and tests pass without later commits (since they are submitted individually).
  3. Run git rebase --continue to move to the next change to edit.

Once all the commits have been edited, the author then re-uploads the changes.

Resolving Merge Conflicts

When resolving merge conflicts it's often useful to rebase a change on origin/master instead of JIRI_HEAD (which is what is checked out when running jiri update). JIRI_HEAD only updates as changes roll through global integration, whereas origin/master contains all submitted changes.