Crate quokka 

A TPX-style external test runner for Buck2, specialized for Rust tests.

Buck2 owns execution and caching. This runner implements buck2’s external test protocol: it serves the TestExecutor service (receiving one ExternalRunnerSpec per target), lists each target’s tests via a cacheable Execute2(Listing), then issues one Execute2(Testing) per discovered test and reports each result. See DESIGN.md for the full design.

Modules

batching

Adaptive batching: collapsing many tiny per-test actions into fewer Execute2 calls to amortize remote-action overhead.

caching

The two distinct caching concepts, kept apart so one bit is not written by several places with no precedence.

cli

Command-line parsing matching buck2’s exact launch contract.

config

db_cli

duration_db

environment

Semantic environment constraints for test execution.

execution

Building Execute2 requests for listing and test actions.

executor_server

The TestExecutor gRPC service the runner serves to buck2.

ids

Typed identifiers and the test-name interner for the scheduler’s hot tables.

listing

Framework-agnostic test discovery: test case data model and ignored-test policy.

orchestrator

Typed client for Buck2’s TestOrchestrator (and DownwardApi) services.

policy

Per-target policy axes derived from labels, each an independent typed value.

proto

Generated gRPC/protobuf bindings for the Buck2 test-runner protocol.

result

Generic (translator-independent) decoding of an Execute2 response, plus the terminal-status, result-identity, log-sink and execution-kind types.

run

The run lifecycle: wire the two channels, serve the executor, drive the scheduler, and — on EVERY exit path, including a scheduler panic — report end_of_test_results before shutting the executor server down.

scheduler

The scheduler: the orchestration that turns intake specs into per-test Execute2 calls and reported results, under bounded, fair concurrency.

spec

The cold, immutable per-target data parsed once from an ExternalRunnerSpec.

translator

Test-framework translators: turning a target’s spec into listing/execution commands and decoding harness output into per-test results.

transport

gRPC transport over a pre-connected duplex stream.

variant

Test execution variant and repetition, the two orthogonal multiplicity axes buck2’s Testing stage exposes (variant, repeat_count).