Skip to content

Latest commit

 

History

History
74 lines (57 loc) · 4.38 KB

1.1.0.md

File metadata and controls

74 lines (57 loc) · 4.38 KB

Lesscode 1.1.0

Prepared 2026-11-11. The tag v1.1.0 is pushed by the owner; the release workflow then builds, signs and publishes the archives as for 1.0.2 (release signing).

Proposed tag message:

Lesscode 1.1.0: one mutation driver for Rust, Python, TypeScript and Java, driven by language packs; it runs a project's own tests against planted faults in copies of the source tree

What is new

  • deconvolute run: plants one fault at a time in copies of your source tree, runs your own build and test commands, restores and verifies the file, and classes every fault as killed, survived, invalid, timeout or error. It never writes to the tree you point it at, and it refuses a work directory inside or around the source. Several copies run in parallel; classed faults are stored and reused.
  • deconvolute analyze and mutate for four languages: Rust, Python, TypeScript and Java, through one engine. Python is new. What the engine knows about a language is a data pack of at most 40 lines.
  • deconvolute admit-pack: checks a language pack (shape, size, node kinds against the grammar, templates, and recorded control cases) before it is used.
  • Stopping a run: Ctrl-C, Ctrl-Break, a termination signal or a closing console kills the project command with everything it started, restores the copy and ends the run as void.

What changed for existing users

  • analyze and mutate no longer use the text-matching rules of 1.0.x, which were bound to this project's experiment specimens and produced nothing on most other code. Receipts are now deconvolution-analysis-v2 and deconvolution-mutation-v2; mutant files are named mutant_NNNN_<operator>.<extension>.
  • analyze --lang python and mutate --lang python return results instead of exit code 2.
  • A source file that does not parse is refused with exit code 3.
  • Library users: the adapters' analyzer and mutant_gen fields are gone; MutantOutcome has a variant NotRun; three functions that reported success without running anything are removed (PureCoreVerifier::compare on the adapters and on NativePureCoreVerifier, NativeMutantGenerator::generate).
  • density, budgets, the pre-push hook, prepare and compare are unchanged.

Measured

  • The driver gives every one of 556 mutants on five real files the class the research prototype gave (R73).
  • Windows binary 13.28 MB (1.0.2: 10.99 MB); two grammars were added.
  • Wall time per fault is the project's own build and test time: about 0.6 s for a Python file, 1.5 s for TypeScript, 5 to 16 s for Java, 5 to 22 s for Rust (R73, R75).

Known limits

  • No command establishes that a rewrite behaves the same. A run measures how sensitive the tests are.
  • Copies must not share a build cache; with Cargo and a global CARGO_TARGET_DIR set "env": {"CARGO_TARGET_DIR": "target"} in the run configuration (driver).
  • The Java and Rust packs do not swap call arguments: measured, 15 of 17 such mutants did not build (R81).
  • A driver that is killed outright cannot clean up at that moment; the copy is restored when next opened.
  • Earlier results of this project that cited "4 of 4 adverse controls killed" are withdrawn as evidence of test sensitivity (R75).

For the owner, before tagging

  1. Tag a commit for which CI is green. The version bump is commit 3455604, whose own CI run failed at the format check of one example file. Tag the commit that switches the argument swap off for Java and Rust (after 3316cbc), once its CI run is green.
  2. git tag -a v1.1.0 -m "<the tag message above>" and git push origin v1.1.0.
  3. Approve the release environment when the workflow asks.
  4. After publication: check the signatures and the attestation as for 1.0.2, and that Zenodo archived it.

Verified after publication (2026-11-11)

Checked from the published assets, downloaded with gh release download:

  • All 12 entries of SHA256SUMS match the downloaded files.
  • SHA256SUMS.sig verifies with the project's public key (the key listed in scripts/release-public-keys.txt), and each of the five archives carries a valid signature.
  • gh attestation verify succeeds for the Windows archive.
  • The Windows binary reports the release's version.
  • Zenodo archived the release: version 1.1.0, DOI 10.5281/zenodo.24135550.- Superseded by 1.1.1 on the same day; its README still shows the line-based density example.