Require 100% region, function, instantiation, line, branch, and MC/DC
coverage in run-coverage.sh. Enable MC/DC instrumentation and rewrite the
two decisions that Clang could not instrument cleanly: the crossed-room
ternary, and the nested boolean under negation in crossedCompleteEndLeOne.
Two mutation rounds plus z3 checks found conditions that an enclosing
dispatch already decides, and two places that re-tested the same split.
* crossed complete: HeadBottom implies NonposBottom, so drop it; inline
the clamp dispatch so EndOne/LongNonpos are not tested twice; after the
failed strict ordering test the non-strict one is free.
* reversed truncation: the dispatch has failed end.offset <= 1, so
BelowEnd does not need to re-test end.offset >= 2; the two TopCrossed
cases split on end.key < last, so test that once; revTruncGeTwoLarge
was a tautological split of TopLarge on isTop begin.
* TopLarge's sentinel re-tested TopLarge even though the caller had
established it; only the sentinel test is needed.
* complete forward: begin is bottom exactly when clampsToBottom, and
bottom <= end always, so the disjunction is just the key comparison.
z3 refutes the negation of each implication. Sweep, golden, corpus, SMT,
extreme, and fuzz pass; fast.cpp stays at 100% line/branch coverage.
Test end.offset <= 1 before revTruncLeOneLow in the reversed crossed
dispatch, and specialize revTruncLeOneLow under that path fact. The
helper drops the now-tautological end.offset conjuncts and the key
comparison implied by a failed optLe.
The crossed truncated reverse benchmark drops from about 67 ns to
about 61 ns. Sweep, corpus, SMT cases, and 100% line and branch
coverage still pass.
Drop the redundant begin.offset split in crossedCompleteOvershoot and
the rE > 0 guard in closedFormForwardSaturated; both are implied by
the call-site invariants.
Add an SMT-derived empty-interval case as the smt_cases CTest and as a
committed fuzz_readable seed. run-fuzz.sh now loads the read-only seed
corpus alongside the writable one.
Rename exhaustive to sweep and document that it is exhaustive only
over its bounded family of instances.
- Correct stale references and headline-path wording in READMEs and Lean docstrings
- Document build steps, comparator requirements, sweep scope, and coverage
- Unify Apache headers across Lean and C++ sources
- Add a bwrap-free axiom audit for the headline theorems
- Add a Gitea workflow running lake build, the axiom audit, C++ tests, and coverage
- Match LLVM coverage tool versions to the compiler and recreate stale build dirs
Drop predicate operands and unreachable arms that dispatch and readable invariants fix, and extend the exhaustive sweep to cover every branch side. Gate coverage on branch coverage as well.
Add an exhaustive readable-instance sweep and a coverage gate requiring
100% line coverage of the summarised implementation. Delete branches that
are unreachable under `Readable`, centralise and assert the 32-bit offset
contract at the public entry points, and stop materialising `n` zero bytes
when deciding `keyAfterN n k < upper`.
Also add the extreme-offset regression test, harden the corpus parsers,
remove unused model helpers, document the literal-total / fast-readable
split, and refresh the validation and benchmark numbers.
- Literal transliteration of `Statement.closedForm` plus an equivalent
summarised implementation
- Borrowed input keys (`std::string_view` / `std::span`) and owned result
keys (`std::string`)
- Golden-vector test, differential/safety fuzzers, and a nanobench
comparison
The literal and summarised implementations agree on 40,000 Lean golden
vectors, 18,971 sorted corpus cases, and the readable fuzzer; the raw fuzzer
runs both implementations under ASan/UBSan. `build/` and the generated
`oracle/golden.txt` are ignored.