Skip to content

fix(nav): tighten SVG geometry contract evidence (#8449) - #8450

Draft
ll7 wants to merge 4 commits into
mainfrom
fix/issue-8314-contract-followups-20260904
Draft

fix(nav): tighten SVG geometry contract evidence (#8449)#8450
ll7 wants to merge 4 commits into
mainfrom
fix/issue-8314-contract-followups-20260904

Conversation

@ll7

@ll7 ll7 commented Sep 4, 2026

Copy link
Copy Markdown
Owner

Summary

Follow up the now-merged PR #8373 by tightening the explicit SVG geometry-contract boundary:
remove overbroad compatibility wording, apply all corrected-mode path translations consistently,
and add focused coverage for parser edge cases, exact legacy state, real-map routes, scenario/cache
propagation, and sub-threshold path translations.

The strict parser and non-SVG fail-closed implementation are already in the base through #8373; this
follow-up makes those guarantees executable and keeps the remaining claims precise.

Linked Issues

Stack / Dependency

What Changed

  • Apply corrected-mode path translations uniformly, including translations below the former
    near-zero threshold used only for paths.
  • Add valid SVG number/list regressions, including 1.e2, whitespace/comma forms, adjacent
    translations, exponent forms, malformed numeric input that must not be partially parsed, and a
    path whose translation components are both below 1e-12.
  • Add canonical SHA-256 checks for every pre-contract MapDefinition state field on all five
    transformed maps under both default and explicit legacy parsing; the additive provenance field
    is intentionally excluded.
  • Add a real corrected bottleneck route assertion and scenario-loader coverage for contract-specific
    cache identity, unknown contract names, and corrected contracts on non-SVG maps.
  • Narrow context and public docstrings from byte-identical/automatic non-pooling claims to the
    supported historical-geometry and explicit-contract guarantees.

Why It Matters

The parent implementation changes the loader contract, so its evidence must distinguish exact
supported behavior from stronger guarantees that the generic downstream aggregation layer does not
enforce. These tests and wording corrections make the compatibility boundary reviewable without
changing frozen results or launching a benchmark campaign.

Research / Evidence Notes

  • Target: diagnostic implementation evidence for explicit legacy/corrected SVG geometry contracts.
  • Comparator: the merged fix(nav): apply ancestor translate transforms with legacy contract (#8314) #8373 implementation, analytic parser fixtures, legacy/corrected map
    outputs, and one authored bottleneck route.
  • Mechanism: complete translation-list parsing in the parent plus contract propagation through map
    loading and its cache; this commit adds regression coverage and consistent path shifting.
  • Failure mode: malformed transform text, invalid non-SVG contract selection, or downstream pooling
    without an explicit compatibility decision.
  • Result: diagnostic-only implementation evidence; no benchmark, ranking, causal, safety, release,
    or paper-facing conclusion.
  • Next empirical action: independent domain-aware review of the exact head, followed separately by
    corrected-contract scenario revalidation if that work is approved.

Validation / Proof

  • Prior exact-head focused navigation/scenario/replay/map suite: 164 passed; the new sub-threshold
    regression file was then validated at the exact PR branch with 65 passed.
  • Ruff check, format check, and git diff --check: passed for the changed implementation/tests.
  • Documentation evidence-integrity and context-note checks: passed on the prior exact head,
    while new hosted checks for this exact head are pending.
  • A full readiness attempt on the prior head passed contract checks and the core lane, then the
    CUDA-safe serial optional lane ran for more than 30 minutes without a terminal result and was
    interrupted cleanly; its termination receipt verified process-group cleanup. The base advanced
    afterward, so that receipt is not current-head proof. Full readiness is not claimed here and this
    PR remains intentionally draft.

Risks / Rollback

  • Legacy remains the default; corrected geometry remains explicit and labeled.
  • This follow-up adds no benchmark execution and changes no frozen artifacts.
  • Rollback is a single commit revert.

Docs / Provenance

  • Updated the SVG geometry context note, context index/README, and public map-contract docstrings.
  • Compatibility digests are anchored to pre-contract parser commit
    23d07d6cf32a6b88027c767d3861e1f00aa1eecc; they prove exact pre-existing-field equivalence,
    not generic whole-object byte serialization including additive provenance.
  • No dataset, checkpoint, campaign output, registry entry, or durable evidence artifact changed.
  • Ignored local output remains temporary and is not part of the PR; no generated artifact is a
    dependency of this draft.

Downstream Propagation

  • MapDefinition.svg_geometry_contract remains the recorded compatibility label.
  • Downstream consumers must partition rows or explicitly establish compatibility before comparing or
    pooling legacy and corrected executions; this PR does not add a generic aggregation guard.
  • Corrected-contract scenario revalidation and any evidence/artifact update remain separate work.

Follow-Up / Residual Scope

  • Complete a clean canonical PR-readiness run, including the optional lane, before converting this
    draft to ready.
  • Obtain independent domain-aware approval on the final exact head before merge or evidence use.
  • Do not replace frozen legacy artifacts or infer planner/safety effects from this implementation PR.

Reviewer Notes

  • Verify the complete-transform-list edge cases, especially translate(1.e2) and malformed
    separators/numbers.
  • Verify the exact legacy projection covers every pre-contract map-state field and excludes only
    additive svg_geometry_contract metadata.
  • Verify the real-map route assertion and the narrowed legacy/aggregation claims.
  • Verify corrected mode changes a path for translate(5e-13,-7e-13) while legacy mode preserves
    the historical path coordinates.

@coderabbitai

coderabbitai Bot commented Sep 4, 2026

Copy link
Copy Markdown
Contributor

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Team

Run ID: 5dabba5e-1476-4b7a-9029-033535471610

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

  • 🔍 Trigger review

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

Comment @coderabbitai help to get the list of available commands.

@github-actions github-actions Bot added the review-bot-auto Managed label that routes code-bearing PRs to CodeRabbit. label Sep 4, 2026
@github-actions

github-actions Bot commented Sep 4, 2026

Copy link
Copy Markdown

🔍 PR Contract Check Summary

Overall Status: 🟢 PASSED

Check Status Description
1. Closes-discipline ✅ PASSED Demand Refs #N for epic issues
2. Closure declaration ✅ PASSED Require Closes/Refs for title issues
3. State-refresh-only ✅ PASSED Reject docs/context state updates
4. Evidence hygiene ✅ PASSED Checks markers and provenance fields
5. Evidence writer usage ✅ PASSED Require the shared marked writer path
6. Successor discipline ✅ PASSED Require successor statement on multi-PR issues
7. Worker-lane label ⚪ None Label PRs from cheap worker lane
8. Placeholder docstring ratchet ✅ PASSED Reject NEW TODO/empty docstrings in added diff lines

ℹ️ Info

  • No worker-lane provenance detected.

This check is mechanized. Please resolve any blockers to pass CI.

@ll7 ll7 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.

Exact-head implementation review

Reviewed PR #8450 at exact head 7ac518f3ff57950f774caf7d06e9ae805b41d05d against
origin/main 5f476625eba9c4ae2fa5770781852b78c865fd8b.

Intended design and evidence boundary

This is a research-adjacent SVG geometry-contract follow-up to merged PR #8373. It narrows
legacy/corrected wording, applies corrected-mode path translations consistently, and adds parser,
real-map route, cache, and scenario-loader coverage. The change is diagnostic implementation
evidence only: it does not replace frozen legacy artifacts, launch a benchmark, or claim planner,
safety, ranking, causal, release, or paper results.

Validation

  • Focused navigation and scenario-loader selection — 102 passed.
  • Ruff 0.16.5 check/format, git diff --check, and clean
    git merge-tree --write-tree origin/main HEAD — passed.
  • The PR body records a broader focused suite and docs/optional-dependency checks; I did not
    independently treat its historical full-readiness attempt as current-head proof.

Findings and disposition

No actionable P0–P3 implementation finding was identified in this exact-head slice. The parser
change is localized, corrected-mode translation is applied only through the corrected extraction
path, and the new cache/contract tests exercise the stated boundary. The remaining admission gaps
are explicit and material: the PR is intentionally draft, the canonical full readiness run is not
clean, independent SVG-geometry/domain approval is absent, and corrected-contract scenario
revalidation remains deferred. Those gaps prevent merge or stronger evidence admission but do not
constitute a new code defect in this review slice.

Decision: hold at draft/diagnostic. Complete a clean canonical readiness run and obtain independent
domain-aware approval before promoting or merging. Do not apply merge-ready.

gate-verdict: blocked @ 7ac518f
pr-metadata: reconciled @ 99d84bed452c7d471f3437086de338379817d4656c02e8b9e9ac48070ff15526

@ll7 ll7 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.

Exact-head implementation review

Reviewed PR #8450 at exact head 7ac518f3ff57950f774caf7d06e9ae805b41d05d against
origin/main 5f476625eba9c4ae2fa5770781852b78c865fd8b.

Intended design and evidence boundary

This is a research-adjacent SVG geometry-contract follow-up to merged PR #8373. It narrows
legacy/corrected wording, applies corrected-mode path translations consistently, and adds parser,
real-map route, cache, and scenario-loader coverage. The change is diagnostic implementation
evidence only: it does not replace frozen legacy artifacts, launch a benchmark, or claim planner,
safety, ranking, causal, release, or paper results.

Validation

  • Focused navigation and scenario-loader selection — 102 passed.
  • Ruff 0.16.5 check/format, git diff --check, and clean
    git merge-tree --write-tree origin/main HEAD — passed.
  • The PR body records a broader focused suite and docs/optional-dependency checks; I did not
    independently treat its historical full-readiness attempt as current-head proof.

Findings and disposition

No actionable P0–P3 implementation finding was identified in this exact-head slice. The parser
change is localized, corrected-mode translation is applied only through the corrected extraction
path, and the new cache/contract tests exercise the stated boundary. The remaining admission gaps
are explicit and material: the PR is intentionally draft, the canonical full readiness run is not
clean, independent SVG-geometry/domain approval is absent, and corrected-contract scenario
revalidation remains deferred. Those gaps prevent merge or stronger evidence admission but do not
constitute a new code defect in this review slice.

Decision: hold at draft/diagnostic. Complete a clean canonical readiness run and obtain independent
domain-aware approval before promoting or merging. Do not apply merge-ready.

gate-verdict: blocked @ 7ac518f
pr-metadata: reconciled @ 99d84bed452c7d471f3437086de338379817d4656c02e8b9e9ac48070ff15526

@ll7

ll7 commented Sep 4, 2026

Copy link
Copy Markdown
Owner Author

Independent targeted verification of head 7ac518f3ff57950f774caf7d06e9ae805b41d05d in a clean detached worktree; comparison ref observed at a2bf33373e5775eacbeed5049c15b987ad5af9df.

  • Shared-environment wrapper, focused geometry/scenario/replay/map tests: 153 passed in 14.18s, exit 0. Files: tests/nav/test_svg_transform_contract.py, tests/training/test_scenario_loader.py, tests/benchmark/test_generated_scenario_replay_adapter.py, tests/test_svg_classic_maps_format.py, tests/maps/test_route_clearance_maps.py.
  • Scoped ruff check and ruff format --check on the three changed source modules and two changed test modules: exit 0; five files already formatted.
  • Inspected the source diff: the functional change removes the near-zero translation shortcut; the documentation narrows legacy/contract claims. No tracked files changed during verification.

This is targeted regression evidence, not full canonical readiness, independent domain approval, or permission to mix historical/corrected campaign rows. The PR remains draft. Existing full-readiness/domain stops are unchanged; no duplicate repair was opened.

@ll7

ll7 commented Sep 4, 2026

Copy link
Copy Markdown
Owner Author

review-claim: codex-goal-autopilot-20260905 @ 7a55403 until 2026-09-05T00:10:20Z

@ll7 ll7 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.

Exact-head implementation review

Reviewed draft PR #8450 at exact head 7a55403246e2c236cd8a3423859ae3cb7f7960ab against current origin/main e7b6f9bfde9f80ca65d62bb1d85c681946e2a3dd in a dedicated linked review worktree.

Findings and disposition

No actionable P0-P3 implementation finding was verified in the scoped review. The diff is limited to the SVG geometry-contract follow-up, scenario-loader contract/cache coverage, documentation wording, and focused tests; it does not replace frozen artifacts or claim benchmark, ranking, causal, safety, release, or paper-facing evidence.

Focused proof at this exact head passed:

  • tests/nav/test_svg_transform_contract.py and tests/training/test_scenario_loader.py: 113 passed;
  • scoped Ruff check and format check: passed;
  • git diff --check origin/main...HEAD: passed;
  • git merge-tree --write-tree origin/main HEAD: conflict-free.

The PR remains intentionally draft and blocked for admission. The live hosted checks still have required jobs pending, the PR body records that the prior optional CUDA-safe serial lane was interrupted without a terminal result, and the required independent domain-aware approval is not present. The body correctly limits its existing domain-approval waiver to implementation smoke and does not treat it as merge or evidence authorization. I did not run a benchmark or the unbounded optional lane locally.

base-policy: ordinary-cas @ 7a55403246e2c236cd8a3423859ae3cb7f7960ab
pr-metadata: reconciled @ f2d61c29cf4b5290765ea7fd72f54b682dab316c7d255542df98c201470f1925
gate-verdict: blocked @ 7a55403246e2c236cd8a3423859ae3cb7f7960ab
merge-ready: no @ 7a55403246e2c236cd8a3423859ae3cb7f7960ab

@ll7

ll7 commented Sep 4, 2026

Copy link
Copy Markdown
Owner Author

review-claim: released @ 7a55403

Exact-head review published as COMMENTED review 5118367130. Disposition: no actionable implementation finding; keep draft and do not merge while hosted checks, independent domain-aware approval, and the clean optional readiness lane remain incomplete.

@ll7

ll7 commented Sep 4, 2026

Copy link
Copy Markdown
Owner Author

Exact-head follow-up status

Verified after the proof update at head
7a55403246e2c236cd8a3423859ae3cb7f7960ab, based on current main
a2bf33373e5775eacbeed5049c15b987ad5af9df:

  • Current-head focused parser/scenario/replay/map suite: 164 passed.
  • Ruff, diff, documentation evidence integrity, context-note freshness, and all
    hosted contract checks pass.
  • Hosted CI is green, including all four fast-feedback shards, changed coverage,
    compatibility, determinism, scenario validation, and CI aggregation.
  • The PR body has been reconciled through the REST writer with metadata digest
    f2d61c29cf4b5290765ea7fd72f54b682dab316c7d255542df98c201470f1925.

The PR remains draft/blocked. The local all-extras environment is available, but a
CUDA-safe serial optional readiness attempt exceeded a bounded 30-minute run without
a result; its process group was cleaned up and the receipt is recorded on #8449.
That receipt predates the latest base refresh, so it is not current-head readiness.
Independent domain-aware approval is also still required. No campaign or frozen
artifact has been changed.

gate-verdict: blocked @ 7a55403
pr-metadata: reconciled @ f2d61c29cf4b5290765ea7fd72f54b682dab316c7d255542df98c201470f1925

@ll7 ll7 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.

Exact-head implementation review

Reviewed PR #8450 at exact head 03998d67c4d7686e81c9f797e12ba0ab2fc9a67c against fresh
origin/main 1cb110f36cbcc96f781c3e5e8f99b13e60a0cb3e in linked worktrees.

Current-head repair

  • The prior P2 coverage gap is fixed with a regression that writes a path translated by
    translate(5e-13,-7e-13), verifies corrected mode applies both components, and verifies legacy
    mode preserves the historical coordinates.
  • No new P0-P2 implementation finding was found in this bounded exact-head review slice.

Validation

  • Exact-branch SVG transform contract suite: 65 passed.
  • Ruff 0.16.5 check/format and git diff --check passed.
  • The PR remains diagnostic implementation evidence only: no benchmark, ranking, causal, safety,
    release, fallback-success, or paper-facing conclusion is made.

The PR should remain draft/blocked. A clean canonical readiness run, independent domain-aware
approval for SVG geometry and historical replay, and corrected-contract scenario revalidation are
still required before merge or evidence use. The prior optional readiness attempt is not current-head
proof.

Decision: hold at draft; the reviewed P2 is repaired, but the declared domain/readiness gates remain
open.

gate-verdict: blocked @ 03998d6
base-policy: ordinary-cas @ 03998d6
pr-metadata: reconciled @ ef612bf9b911b596d80bb53f47356f3624315ddab043fdfdb2f566e5155aa835

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

review-bot-auto Managed label that routes code-bearing PRs to CodeRabbit.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant