Debugging Server project
Debugging with vscode
Using a debugger sometimes is much easier than trying to just print things to understand a bug,for the server project we have a configuration that makes it easy for everyone to run in debug mode inside vscode.
In the debug screen you can select "Server" and hit the play button to start the process.
Before running the server there is a task called transpile-server that runs yarn tsc to make sure that thelatest changes in the package/server are being executed.
In a optimal solution we would be doing this transpile-server task in the tasks.json file(inside the tasks.json we can use a type npm for the task that is more ergonomic)
If there are any new environment variables that need to be included before execution, there is a place where they can be added:
Something that we could do to improve the experience is adding a build process to other projects that are used by the server project,but this still can be done with other terminals open.
Modifying the configuration:
To modify the configuration, for now, it is all in the joplin.code-workspace file, but if needed we could also breakthe configuration into two files, one for the launch.json and other for the tasks.json.
References:
Launch option to vscode workspaces
More images:
Running debug commands
When running in development mode, several debug commands can be run by sending requests to /api/debug. These include:
- The
populateDatabasecommand adds content to the database, creating test users with initial data.- Among other things, this allows testing how Joplin Server handles a large number of users, items, and changes. A larger
sizeparameter creates more items.sizecan be either 1, 2, or 3. - Example:
curl --data '{"action": "populateDatabase", "size": 2}' -H 'Content-Type: application/json' http://localhost:22300/api/debug.
- Among other things, this allows testing how Joplin Server handles a large number of users, items, and changes. A larger
- The
benchmarkDeltaPerformancecommand tests the performance ofmodels.change().deltaby callingdeltamultiple times for each user account. Data is saved inpackages/server/delta-perf.csv.ChangeModel.deltais called during sync and has historically been a performance bottleneck.- Example:
curl --data '{"action":"benchmarkDeltaPerformance"}' -H 'Content-Type: application/json' http://localhost:22300/api/debug.
Fuzzing
The sync fuzzer looks for sync bugs. It works by:
- Starting an instance of Joplin Server in development mode.
- Starting instances of the CLI app and connecting them to the server.
- Performing pseudorandom actions (e.g. "create note", "delete note", "sync").
The fuzzer maintains a model of the expected state of the CLI apps. When the actual state of the CLI apps no longer matches this model, the fuzzer stops.
To start the fuzzer, run yarn syncFuzzer start from the main Joplin workspace directory. See yarn syncFuzzer start --help for more information.
Note: If you encounter an "unauthorized" error, it may be necessary to set the FUZZER_SERVER_ADMIN_PASSWORD environment variable before running the fuzzer.
Fuzzing with breakpoints
To pause the fuzzer at breakpoints in the client/server logic, use a "JavaScript Debug Terminal" in VS Code:
- Open a debug terminal in VS Code (command palette > "Debug: JavaScript Debug Terminal").
- Start the sync fuzzer (
yarn syncFuzzer start).
When debugging, the --setup=<path to setup file>, --snapshot-after=<steps> and --restore-from-snapshot options can be particularly helpful (see yarn syncFuzzer start --help).