Skip to content

docs(#536): record CC-536-01-01 Completion implementation status after #562 - #564

Merged
raylei50653 merged 2 commits into
mainfrom
docs/536-completion-writeback
Oct 10, 2026
Merged

raylei50653 merged 2 commits into
mainfrom
docs/536-completion-writeback

Conversation

@raylei50653

Copy link
Copy Markdown
Owner

Writes back the Completion implementation (#562, merge 781472dc, head 789eace0 with the #563 republication) into the contract doc. Docs only, single file. Same shape as #560 for CC-536-08-02.

  • CC-536-01-01 → implemented (source level). New status block keeps apart:
    • CPU in CI: the 181-check protocol test and the reader negative controls.
    • Local only: the real-binary CLI tests and the reader↔writer cross-check, run against binaries rebuilt at 789eace0.
    • Local mutation: 12 mutants.
    • Local GPU: at 006007d3, 7/7 EXACT with MOT bytes unchanged.
    • Unverified: packaged Completion first, plus power loss, network or WSL filesystems, F3, and probe ≠ equivalence.
    • Known limits: including the non-blocking P3 reader gap on a corrupt track_report.json, routed to architecture(G): validate architecture diagrams, enforce contracts and hand coverage to #509 #541.
  • Packaged Completion is explicitly not claimed. entrypoint_pin.json still pins the PR-C2 binary, and the re-pin is a separate slice that is not yet authorised.
  • Pointers updated: status wording (§0), diagram legend (N-T1 implemented), G1 row, card intro and Evidence line, §2.2 item 6 (baseline kept, with a pointer to the current behaviour), §5 rows N-R8 and N-T1, and §6 item 2.
  • Unchanged: the design decision, the diagram source, and CC-536-01-02, which stays accepted target.

Local scripts/pre_push.sh: all checks passed (5441 passed).

🤖 Generated with Claude Code

Writes back the Completion implementation (#562, merge 781472d, head
789eace with the #563 republication) into the contract doc: implemented
scope, CPU/CI, local-only real-binary, local mutation and local GPU evidence
kept apart, unverified items (packaged Completion first: entrypoint_pin.json
still pins the PR-C2 binary, so only source-level Completion is claimed) and
implementation-level known limits, including the non-blocking reader gap on
a corrupt track_report.json routed to #541. Only CC-536-01-01 changes state;
CC-536-01-02 stays accepted target. The design decision and the diagram are
unchanged (legend notes N-T1 is implemented); §2.2 item 6 points to the
current behaviour.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…n as sole journal writer

Address #564 review: the status no longer says "target interfaces 1–6 all",
since step 4 references the CUDA-free Preflight (N-T2), which belongs to
CC-536-01-02 and is still an accepted target. The State writer line now names
RunCompletion as the only journal writer (run_sequences commits via run(),
main records failure via fail()).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@raylei50653
raylei50653 merged commit ef558ee into main Oct 10, 2026
7 checks passed
@raylei50653
raylei50653 deleted the docs/536-completion-writeback branch November 13, 2026 11:09
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant