Declare forms capability for fixture 37 and verify Jetpack materialization

[chubes4]

Problem

37-art-gallery-exhibition is tagged has-form but declares no forms capability. The fixture matrix therefore normalizes capabilities: [], never provisions Jetpack, never activates the contact-form module, and never runs provider readiness. Nine otherwise mappable forms remain html_form_fallback: seven repeated footer newsletter forms plus the contact and visit forms.

Homeboy evidence: run 5aedff96-df46-466e-b2db-e7e84e88b347.

Work

• Add the forms capability to the fixture metadata.
• Run the fixture through Jetpack provisioning/readiness.
• Verify all supported controls materialize into native Jetpack form blocks/markup.
• Confirm zero visual regressions and editor-valid output across the bounded surface matrix.
• File a mapper defect only for controls or shell semantics Jetpack actually declines.

The free-form has-form tag remains useful for querying but must not substitute for capability declaration.

[chubes4]

Post-merge matrix run 01629418-9c5f-4875-82f0-203c59db7daf now preserves editor and visual evidence after wp-codebox#2125. It exposes two form-materialization gaps:

• SSI emits malformed Jetpack persisted block trees; tracked in Emit editor-valid Jetpack form block trees static-site-importer#737.
• generic/forms/v1 retains flat form/control metadata but drops source presentation structure between them. Fixture 37 relies on wrapper classes including .row-2 and .field; SSI cannot reconstruct those rows from the current contract, so all seven surfaces mismatch (aligned differing-pixel ratios 3.58% to 58.05%).

The upstream repair should carry a generic, bounded form layout/group structure or equivalent source presentation metadata through html_form_fallback and ArtifactCompiler, without adding Jetpack semantics to Blocks Engine. This remains part of #700 acceptance because declaring the capability alone does not preserve the fixture presentation.

[chubes4]

Correction to the earlier fixture setup: capabilities: [forms] must not be required to provision Jetpack. SSI already infers forms from arbitrary HTML and materializes its configured provider.

I removed the declaration in commit 7e65343c and removed capability-driven setup from the SSI fixture matrix in 632d12ea (SSI #738). Homeboy run 0e7f3aab-18b7-4746-a428-1b0f6c8084a2 used fixture 37 without a forms capability and no Jetpack recipe setup. The import passed and all Jetpack blocks validated; only the unrelated core/separator style mismatch remained.

The fixture capability change should therefore be dropped from #700. Fixture metadata may describe expected coverage, but it must not be an operational dependency declaration.

[chubes4]

Fresh evidence supersedes the original explicit-capability premise. SSI PR #740 intentionally infers provider lifecycle from runtime declarations while fixture metadata remains capability-free. That path now provisions Jetpack and materializes the forms without adding forms to fixture metadata.

Eight-surface proof against SSI PR #748 and Blocks Engine 0cd65716: Homeboy run 4610bb1b-6a98-40fb-b2b6-bed00bc1fba3.

Observed:

• All seven captured pages now contain real visual-compare evidence, correctly recognized after SSI #747/#748.
• Every surface fails visual parity: front 7.29%, art-fairs 4.03%, artists 33.81%, contact 16.87%, current-exhibition 58.28%, exhibitions 11.17%, visit 20.94%.
• Dominant normalized cause: restyle_geometry (23,413 classified pixels).
• The pattern appears only after inferred Jetpack materialization; Keep separator geometry block-valid #704 recorded zero visual mismatches before provider activation.
• The generic forms/v1 entity contract flattens controls and drops source wrapper topology/classes. SSI receives the outer .form class but cannot reconstruct .row-2 > .field, so source layout selectors no longer match provider markup.
• A separate core/column max-width validity regression is being tracked independently.

Recommended remaining contract for this issue:

1. Keep fixture metadata capability-free and verify inferred provider provisioning.
2. Preserve bounded, provider-agnostic form wrapper topology in generic form metadata.
3. Let the Jetpack adapter map that topology to valid provider blocks/wrappers without fixture-specific selectors.
4. Rerun all eight surfaces and require zero invalid blocks and zero visual mismatches.

Evidence index: https://homeboy-artifacts-tunnel.dev.chubes.net/4610bb1b-6a98-40fb-b2b6-bed00bc1fba3/ (the tunnel was temporarily returning 404 during inspection; the durable Homeboy run remains authoritative).

AI assistance: OpenAI GPT-5.6 Sol via OpenCode/Homeboy correlated the runtime declaration path, flattened forms/v1 shape, and surface-level parity evidence. Chris owns the conclusion.