Solve 27-university-department fixture against the released corpus gate

[chubes4]

Goal

Make fixtures/websites/27-university-department a solved_candidate using the released milestone stack, then prove fixtures/solved/15-saas remains solved in the target-plus-solved promotion lane.

Baseline

• PHP Transformer: php-transformer-v0.2.7
• Static Site Importer: v1.3.0
• Promotion gate: Automattic/static-site-importer#660
• Prior six-fixture evidence measured a 55.38% aligned visual mismatch. That evidence predates the merged navigation gap repair in Preserve source list gap on converted core/navigation wrappers #617, so establish a fresh target-only baseline before selecting repairs.

Acceptance criteria

1. Run the SSI target lane for 27-university-department on current accepted refs and capture editor, block-validity, and exact visual evidence.
2. Fix generic owning-layer causes only; no fixture-specific selectors or compensating output patches.
3. Reach solved_candidate with exact visual parity, valid native blocks, and no unaccepted runtime/HTML islands.
4. Run the promotion lane selecting 27-university-department plus every fixture under fixtures/solved/.
5. Confirm both the target and 15-saas retain solved_candidate; any solved_regression blocks promotion.
6. Promote fixture 27 into fixtures/solved/ only from accepted gate evidence.

Compatibility

Any generic transformer behavior change must document downstream compatibility impact and extension-path implications.

AI assistance

• AI assistance: Yes
• Tool(s): OpenCode with OpenAI GPT-5.6 Sol
• Used for: Historical evidence review, release-baseline coordination, and issue drafting.

[chubes4]

Fresh released-stack baseline from Homeboy run b46a970c-d5a6-41df-9b69-bdcbd584390b:

• Raw full-render mismatch: 3,544,399 / 5,168,640 pixels (68.58%); aligned comparison reported 67.33%. Candidate height is 342px taller.
• The first measured divergence is the header: source 1280x133.23, candidate 1280x209.47 (+76.24px). Main content consequently starts +110.04px lower and grows +360.58px.
• Utility copy changes from a source block-level span with zero margin to core/paragraph with 12.48px bottom margin, inflating the utility row by 12.47px.
• Primary navigation changes from source display:block, 777.69x50.11 to native Navigation wrapper display:flex; flex-wrap:wrap; gap:9px, 663.72x68.38. The generated inner ul does receive the source .primary-nav ul { display:flex; gap:0 }; the remaining divergence comes from canonical wrapper geometry and reduced available width, not a missing nav type-selector rewrite.
• Dominant classification is restyle_geometry (1,539,875 pixels).

The repair should remain generic: preserve source geometry when inline semantics become block wrappers and when core/navigation adds canonical wrapper layers. It must be proven with computed header/nav measurements, the full 27-university-department target lane, composer test, and the solved 15-saas promotion gate. Local pulled artifacts are operator evidence; reviewer-facing results will be translated into PR measurements and reproducible repo commands.

[chubes4]

An isolated generic candidate is prepared while Homeboy runner recovery is blocked by Extra-Chill/homeboy#9246. The candidate addresses the first verified divergence: standalone inline flex/grid items canonicalized as native paragraphs now preserve resolved source margins, or receive a native zero-margin baseline when the source has none. Author-selector-owned structural/ID margins remain on their projected semantic wrapper without a competing paragraph margin.

Deterministic verification passes: composer --working-dir=php-transformer test, all 244 parity fixtures, and git diff --check. This is not yet claimed as the fixture solution: no browser rerun has occurred, and primary-navigation/brand wrapper geometry may remain. Homeboy must ingest the candidate after control-plane recovery, rerun 27-university-department, then run the solved 15-saas promotion gate before PR finalization.