Skip to content

V1_GATE.md

Latest commit

 

History

History
110 lines (93 loc) · 6.24 KB

V1_GATE.md

File metadata and controls

110 lines (93 loc) · 6.24 KB
Raw

SpliceCraft v1.0.0 acceptance gate

The 0.x line is a sustained beta with weekly-to-daily releases. The 1.0.0 tag carries different promises — API stability, semantic versioning, longer support window — and so the gate to cut it is deliberately strict.

Maintainer-approval gate. Meeting every criterion below does NOT auto-clear the gate. The 1.0.0 tag is cut only with explicit maintainer approval, even if everything on this list is green. This file is the checklist, not the trigger.

Target date

2026-05-24 is the working deadline. It is not a commit. If criteria slip, the date slips.

Hard gates (must all be green)

# Criterion How we check it
1 Zero open bug-labelled issues marked severity:high or severity:critical gh issue list --label "bug,severity:high" --state open returns empty
2 Full test suite green on every Python version in CI matrix (3.10–3.13) gh run list --workflow=test.yml --branch=master --limit=1 shows success
3 All 41 sacred invariants (CLAUDE.md) have at least one regression test manual audit + invariant-to-test crossreference grep
4 Pyright strict passes on splicecraft.py and splicecraft_cli.py pyright exits 0
5 ruff check passes ruff check . exits 0
6 Coverage ≥ 80% on biology primitives (_scan_restriction_sites, _rc, _iupac_pattern, _translate_cds, _feat_len, _bp_in, _rebuild_record_with_edit) coverage report --include="splicecraft.py" --fail-under=80 after the targeted run
7 All 8 terminals in RELEASE_CHECKLIST.md smoke matrix pass physical sign-off in the checklist for the release candidate
8 Pre-update snapshot + restore round-trip verified on a real user data dir manual run of splicecraft update --restore-pre-update latest
9 Bundled .dna round-trip verified against the commercial editor's free viewer manual run on AB303066.dna
10 README screenshot is current within the release candidate visual diff against the current splicecraftScreenshot.png
11 CHANGELOG entry for 1.0.0 reads as a coherent release narrative — not a commit dump maintainer review
12 Public docs site published (mkdocs on GitHub Pages, or equivalent) https://binomica-labs.github.io/SpliceCraft/ resolves
13 SECURITY.md disclosure channel tested end-to-end maintainer self-test
14 No TODO / FIXME / HACK markers in splicecraft.py or splicecraft_cli.py `grep -nE "TODO

Soft gates (strongly preferred, not blocking)

# Criterion Why
S1 One non-maintainer code reviewer has signed off Bus-factor mitigation
S2 At least one merged community PR in the 0.x history Sanity check on the contribution path
S3 _run_update_subcommand (currently 532 lines) and the other 200+-line functions are below 200 lines each Reviewability
S4 Conda-forge / bioconda recipe has been accepted upstream Reach beyond PyPI
S5 Performance regression suite (tests/test_performance.py) has CI-asserted baselines, not just smoke-runs Catch silent perf cliffs
S6 PlasmidApp (currently 156 methods) factor-OR-document decision resolved 2026-05-17: mixin extraction declined as cosmetic — controllers cut through call chains, not around them. See PlasmidApp.__doc__ for the full rationale. v2.x viewmodel layer is the proper-decoupling project. Long-term maintainability
S7 Issue templates have been used at least once by a real reporter Validates the channel
S8 Conda recipe has an open PR or merged PR in bioconda-recipes Documents the canonical-copy promise

What v1.0.0 commits us to

These are the post-1.0.0 promises a maintainer should be prepared to keep, and the reason the gate is strict:

  • Semantic versioning. Breaking changes require a major bump (2.0.0). Behaviour changes inside a minor bump must be backward-compatible at the file-format, agent-API, and CLI levels.
  • Two-version support window. Bugs filed against 1.x.y get a fix in the latest 1.x if x is current-1 or newer. Older 1.x lines are upgrade-only.
  • Data format stability. Schema versions for plasmid_library.json, collections.json, parts_bin.json, primers.json, feature_library.json, etc. require migrators for every bump (the _ENTRY_MIGRATIONS registry is the lever; invariant #36). A user dropping a 0.x data dir into a 1.x install must boot cleanly and never lose entries.
  • Agent-API surface stability. No endpoint removal without a 1.x minor-bump deprecation window. New endpoints are additive.
  • CLI flag stability. No flag rename or removal without the same deprecation window. New flags are additive.

What v1.0.0 does NOT commit us to

  • Unicode rendering stability. Terminal emulators ship glyph changes constantly; the braille-dot map will shift visually with the user's fonts and that is by design.
  • Performance. We document budgets (< 5 s first render on 5 MB, < 33 ms/frame) but a 1.x minor release may slip these if the trade-off is clearly worth it. Regressions WILL be flagged in release notes.
  • Optional-dependency behaviour. primer3-py and pyhmmer are required today; should either project become unmaintained, we reserve the right to ship a 1.x that vendors a fallback or reduces functionality, documented in release notes.

Tracking

Current state (snapshot at the time of last edit; re-run before cutting the release candidate):

Gate State
Hard #1 verify before RC
Hard #2 verify before RC
Hard #3 verify before RC
Hard #4 verify before RC
Hard #5 verify before RC (new — added 2026-05-17)
Hard #6 verify before RC
Hard #7 verify before RC
Hard #8 verify before RC
Hard #9 verify before RC
Hard #10 verify before RC
Hard #11 verify before RC
Hard #12 verify before RC (new — added 2026-05-17)
Hard #13 verify before RC
Hard #14 green (0 markers as of 2026-05-17)

Each "verify before RC" should be re-checked and flipped before proposing the 1.0.0 tag. The maintainer-approval gate sits above all of these.