For developers #

go test ./...                                   # unit, end-to-end and golden tests
go test -race ./...
go test ./internal/api -run Golden -update      # refresh output snapshots after an intentional change
go run golang.org/x/vuln/cmd/govulncheck@v1.8.0 ./...

• Golden tests compare the output with 68 real Twitter API responses from 2010 (testdata/reference, with sources listed in SOURCES.md). Every field must exist with the right type, and XML elements must appear in the right order. mockingbird's exact output is also saved in testdata/golden. To refresh the references, run python3 scripts/fetch_wayback.py && python3 scripts/extract_refs.py.
• Live tests run against a real Bluesky account when you provide one:

  MB_TEST_HANDLE=you.bsky.social MB_TEST_APP_PASSWORD=xxxx-xxxx-xxxx-xxxx go test ./internal/integration -v
  

  Add MB_TEST_ALLOW_WRITES=1 to also post, like, reply and delete, or MB_TEST_DM_APP_PASSWORD to test messages.
• Load test: mockingbird loadtest -users 3000 -rps 100 -duration 2m simulates thousands of old apps polling a pretend Bluesky server. It reports the delay mockingbird adds, memory use and data sent. It also works inside the image: docker run --rm mockingbird:local loadtest.

Code layout