test-run

Run Tests

Context

Always use tail when running tests to preserve context. Run all tests piping the entire output to .agents/tmp/scratch.txt, use this for all commands, so that multiple greps can be applied without rerunning the tests.

Hanging Tests

If instructed, or you suspect, that some tests may hang, use timeout to catch these. Run an exponential backoff timeout, starting with two minutes, to catch hanging tests once you find a hanging test, shorten the timeout as much as possible to speed iteration cycles. When debugging hanging tests always pass the log-runs flag to detect the tests that dont finish, ie: timeout 120 cargo test -p beet_core --lib -- --log-runs --test-threads=1. A test logs RUNS as it starts and the runner starts at most --test-threads at once, so with one the last RUNS line is the test that didnt complete.

Compile Errors

these sweeping tests perform large compilations and sometimes the mold linker trips up. try running again, and if it still trips up sometimes incrementing the workspace and crates version, ie find-replace Cargo.toml 0.0.9-dev.6, ..dev.7, will unstuck it. A classic linker error is: help: you can increase rustc's stack size by setting RUST_MIN_STACK=.. this can usually be fixed by running again, and getting a little further down the compilation step.

Instructions

Run just test-core If you encounter an error, isolate it and run again, ie if the error is in beet_core lib: cargo test -p beet_core --all-features --lib -- test_name Fix the error using a subagent, then run the crate again: cargo test -p beet_core --all-features When thats clear, run the full suite again just test-core | tail After the first complete pass, run fully again, checking for warnings. Fix any warnings encountered

Success

Success means that just test-core passes without warnings. Every warning counts, including ones in code you did not touch or another agent just committed: fix it, do not report it as pre-existing. Upon completion provide a comprehensive summary of what was changed.