simon/jolt

Yogthos 9a273e71cd Move the Chez test oracle off Janet

The unit tests in test/chez/_*.janet now drive bin/joltc (the zero-Janet
spine) and judge against baked expected values instead of a live build/jolt
run. Ten of them captured the oracle from build/jolt per case; their values
are now literals (one env-dependent javastatic case became a predicate so it
stays portable). The rest already had literal expecteds with a redundant
build/jolt sanity check, now dropped.

Retire emit-test/emit-parity/reader-parity: they compared the Chez/Clojure
path against a live Janet evaluation, emitter, or reader. That migration check
is done, and run-corpus-zero-janet (Chez analyzer vs the JVM corpus) plus
certify.clj cover correctness now.

Rewrite the README for the current zero-Janet gate.

jolt-5oci

2026-06-21 09:48:49 -04:00

2.9 KiB

Raw Blame History

Chez test harness

The correctness gate for jolt on Chez (epic jolt-cf1q). Correctness is judged against the JVM-sourced conformance spec — no Janet decides pass/fail. The spec itself lives in test/conformance/ (see its SPEC.md); this directory holds the Chez runners and unit tests.

The spec corpus

corpus.edn — the contract: ~2920 rows {:suite :label :expected :actual}, with :expected sourced from reference JVM Clojure. Valid as both EDN and Janet data. Generated by test/conformance/regen-corpus.clj (the answers) over the case list from extract-corpus.janet (the :actual strings, pulled from test/spec/*.janet and conformance-test.janet).
extract-corpus.janet — re-derives the case list when spec rows change. Writes only when JOLT_EXTRACT_CORPUS_OUT is set, so a gate run never clobbers the JVM-sourced corpus.edn. After re-deriving, re-source the answers with regen-corpus.clj.

The standing gate

run-corpus-zero-janet.janet runs every corpus case through the zero-Janet spine: read → analyze → IR → emit → eval, all on Chez (the analyzer is jolt.analyzer cross-compiled to Scheme over host-contract.ss). Each result is compared by value-equality against the JVM :expected. A known-fail allowlist covers cases jolt can't match because Chez has no JVM host (Java classes, arrays, BigDecimal, opaque host-object printers, …); the gate fails only on a NEW divergence or if the pass count drops below the floor.

JOLT_CHEZ_ZEROJANET_CORPUS=1 janet test/chez/run-corpus-zero-janet.janet
JOLT_CORPUS_LIMIT=200 …            # every-Nth stride, fast iteration

Floor: 2678 (JOLT_CHEZ_ZJ_FLOOR overrides). Raise it as host gaps close.

Unit tests

_*.janet drive bin/joltc (the pure-Chez CLI; JOLT_BIN overrides) one subprocess per case and compare the last stdout line against a baked expected value. :throws asserts a non-zero exit. Each file targets one area — _type/_class/_seqpred (taxonomy + predicates), _str/_strns/_reader (strings + reader), _io/_ioreader/_insttime/_javastatic (host interop), _dynbind/_var_meta/_ns (vars + namespaces), _atomwatch/_walk/_dotform /_stdlib (refs, walk, dot-forms, stdlib). Run one with janet test/chez/_type.janet.

Self-host gates

spine-test.janet — the zero-Janet compile/eval spine end to end.
cli-test.janet — bin/joltc -e behavior.
bootstrap-test.janet — host/chez/bootstrap.ss rebuilds the prelude + image from source on Chez and matches the checked-in seed (host/chez/seed/).
fixpoint-test.janet — the on-Chez compiler reproduces itself (stage2 == stage3, prelude pstage3 == pstage4).

Bench

bench-pipeline.janet (opt-in via JOLT_CHEZ_BENCH=1) times fib + mandelbrot through the real pipeline against the spike ceiling. Skipped in the gate.

All Chez runners skip cleanly when chez is not on PATH.

2.9 KiB Raw Blame History