Fix validation crashing on empty enum in 2019-09 and 2020-12

[jviotti]

Signed-off-by: Juan Cruz Viotti jv@jviotti.com

[cubic-dev-ai]

No issues found across 20 files

[augmentcode]

🤖 Augment PR Summary

Summary: Updates the vendored Blaze engine to prevent validation crashes when an enum keyword is present but empty (notably affecting newer JSON Schema drafts).

Changes:

• Bumps the vendored vendor/blaze dependency and its embedded JSON-Schema-Test-Suite revision.
• Adds a new Blaze documentation component (CMake option, install config wiring) plus a schema describing the documentation JSON format.
• Introduces new canonicalizer rules (InlineSingleUseRef, AllOfMergeCompatibleBranches) and adjusts existing canonicalization behavior around allOf, refs, and type inference.
• Improves type inference in in-place applicators (including inferring types from non-empty enum siblings and some $ref targets).
• Adjusts alterschema transformer reference-rewrite logic to better handle relocated reference origins after transforms.
• Fixes empty-enum validation by compiling it into an unconditional failure and adds a clearer evaluator message for that case.

Technical Notes: The empty-enum case now produces an AssertionFail instruction and a dedicated human-readable message during evaluation.

_{🤖 Was this summary useful? React with 👍 or 👎}

[augmentcode]

Review completed. 2 suggestions posted.

Comment augment review to trigger a new review at any time.

[augmentcode]

vendor/blaze/src/documentation/documentation.cc:595: resolve_ref returns the referenced schema and discards any sibling annotations on the $ref wrapper (e.g., title/description), which means walk_properties/walk_pattern_properties may omit property-level documentation metadata when schemas are written as $ref + annotations.

Severity: medium

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

[augmentcode]

vendor/blaze/src/documentation/documentation.cc:1343-1345: VisitedEntry.identifier is set to next_identifier before walk_schema() consumes that value for the table identifier and then increments again for the first emitted row. If recursiveRef.identifier is meant to point at a target row identifier (as implied by the HTML rendering/target-collection logic), this looks like an off-by-one that could break recursive reference linking.

Severity: medium

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}