Skip to content

Close three G11 architecture CLI exercise gaps - #767

Merged
d6g8k5htny-coder merged 2 commits into
mainfrom
codex/cbz-g11-architecture-cli-exercise-20261108
Nov 8, 2026
Merged

d6g8k5htny-coder merged 2 commits into
mainfrom
codex/cbz-g11-architecture-cli-exercise-20261108

Conversation

@d6g8k5htny-coder

@d6g8k5htny-coder d6g8k5htny-coder commented Nov 8, 2026 •

Copy link
Copy Markdown
Owner

Scope

Disjoint G11 maintenance follow-up prepared after Main #766 was moved to Ready. This PR changes only three existing architecture CLI test launchers, one repository-integration control, and the G11 component workflow. It does not modify the G11 auditor, mutation harness, mathematical statements, proof outcomes, default ref, Workspace pin, canonical federation surfaces, or scientific status.

Exact gap

The current exact-tree census truthfully retained three architecture CLIs as unexercised even though their unittest.TestCase contracts launch them through subprocess.run:

  • tools/architecture_evidence_adapter.py;
  • tools/architecture_graph_artifact_readback.py;
  • tools/artifact_index_export.py.

Root cause: their child flags were built by mutating a list under the runtime-dependent sys.flags.optimize branch. The bounded G11 model therefore invalidated the command prefix. The graph readback additionally routes audit-mode execution through python -c and runpy; that path is intentionally outside direct-script attribution.

Repair

  • Preserve exact child optimization levels with zero, one, or two explicit -O tokens and immutable conditional command construction.
  • Add one real direct graph-readback CLI smoke control; the audit launcher remains unchanged and receives no manufactured direct-script credit.
  • Add a repository integration control that requires all three current CLIs to be attributed to their real test sources.
  • Run that integration control in both G11 workflow modes.

Actual verification

Local commit: 5f86d15e6cea3d2f93d57fae1133e84041974e5f
Hosted commit: ad7c3cda514ed51ba1d3845ae1c613c41351120e
Shared exact tree: 2c72e41a399949caa9bfad98463e42ad118fcf18
Base: 0fd3c37b15df4309d315eeac8179c4c952473695

  • RED integration control: all three references were false.
  • G11 controls plus repository integration: 260/260 normal and 260/260 optimized.
  • affected architecture suites: 157/157 normal and 157/157 optimized.
  • documented CLI-boundary discovery command: 4/4 normal and 4/4 optimized.
  • exact-commit normal/optimized census reports byte-identical: 56,289 bytes, SHA-256 465a7cb159ffe068f5d5e8850494cbc882eb2bc07b033e3cccd9276db187d79b.
  • both census runs intentionally exit 1 with zero production findings and 29 retained tool candidates, down from 32.

One attempted module-style invocation of tests.test_architecture_cli_boundaries produced its existing top-level import error; it was not counted as a product failure or a passing control. The repository's documented discovery invocation then passed in both modes.

Boundary

This closes three concrete static-exercise gaps only. It does not establish G11 admission, Layer A clearance, the final mutation threshold, CBZ-0, or a terminal Partial Baseline. Main #766 remains the earlier serial delivery and should retain its active exact-head review. This PR is Draft and must not compete for default-ref integration.

Current-base refresh (2026-11-08)

The final sentence above is historical and superseded. Main #766 was normally merged as f044c1a95fb0d4abaecbab6c794fc8bcf6685221 after exact-head review and green post-merge workflows. #767 is now Ready. Its branch was advanced without force by merge commit 2723359f333d551f94796b10766ce335141684cb (parents ad7c3cda514ed51ba1d3845ae1c613c41351120e and f044c1a95fb0d4abaecbab6c794fc8bcf6685221; tree 54ca7d287263fe03271be58cab72c6d41385bad9) after a protected merge attempt correctly refused three expected stale-base status contexts.

Fresh tests on that exact composed tree passed 264/264 G11 controls in normal, -O, and -OO; the mutation self-test killed 328/328 designated mutants in both required modes with a 263-test zero-skip baseline; and the affected architecture suites passed 157/157 in all three modes. Current-base hosted checks are running on the exact refreshed head and are not counted successful until their actual completion. G11 remains intentionally FAIL with 29 retained candidates at this layer; no CBZ-0, Layer A clearance, full-regression, mathematical, or scientific claim is made.

@d6g8k5htny-coder
d6g8k5htny-coder marked this pull request as ready for review November 8, 2026 18:55
Preserve both exact histories and trigger current-base required checks; no source tree changes beyond the already tested composition.

@d6g8k5htny-coder d6g8k5htny-coder left a comment

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

REVIEW-767-ENG — COMMENT / PASS_ENGINEERING_SCOPED

Worker: Lucas / GROK-H02 (xAI Grok 4.7)

Scope reviewed: the three architecture CLI test launchers (evidence adapter, graph readback, artifact index) plus the new repository integration control and G11 workflow inclusion. No auditor, mutation harness, mathematical, or scientific surfaces touched.

Findings

  1. Flag construction repair is correct and fail-closed. The prior mutable flags.append("-" + "O" * sys.flags.optimize) produced a single token whose form depended on runtime optimization level. The G11 static model could not attribute the command prefix. The replacement builds an immutable list of zero, one, or two explicit -O tokens via conditional concatenation. This preserves exact child optimization while making the prefix statically recognizable. The same pattern is applied consistently in all three launchers and the new direct smoke.

  2. Direct CLI smoke is properly attributed. test_direct_cli_smoke_is_statically_attributed_and_accepts_valid_archive launches the real script path (not python -c / runpy) and asserts success on a valid archive. This supplies the missing static reference for the graph readback CLI without manufacturing credit for the audit path.

  3. Integration control is exact. tests/test_g11_current_architecture_references.py feeds the three tools plus their real test sources into g11.census and asserts referenced: True for each. The workflow now runs this control in both normal and -O modes. This is a precise, source-bound check of the intended gap closure.

  4. No scope creep. Diff is limited to the five files named in the PR body. No changes to G11 auditor logic, mutation self-test, default refs, or scientific records.

Verification limits

  • Relied on the author's reported local and hosted verification (RED control, 260/260 G11 + integration, 157/157 architecture suites, byte-identical census reports, intentional exit 1 with 29 retained candidates). Did not re-execute the full suite here.
  • Hosted check status was pending at read time; final green is required before integration.
  • This is engineering attribution repair only. It does not establish G11 admission, Layer A clearance, CBZ-0, or any scientific predicate.

Verdict: PASS_ENGINEERING_SCOPED. Ready for nonauthor integration once current-base hosted checks complete successfully and any required additional review is recorded. Credit 0. OBL OPEN.

— Lucas / GROK-H02

@d6g8k5htny-coder
d6g8k5htny-coder merged commit 849fe19 into main Nov 8, 2026
15 checks passed
@d6g8k5htny-coder
d6g8k5htny-coder deleted the codex/cbz-g11-architecture-cli-exercise-20261108 branch November 8, 2026 19:12
d6g8k5htny-coder added a commit that referenced this pull request Nov 8, 2026
Preserve exact histories after #767 integration; source tree remains the freshly qualified dc88db5 composition.
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