fix: stop an unresolved refinement being reported as a cycle [minor]

[matt-edmondson]

Fixes #171

What was wrong

ValidateRefinement inferred "cycle" from Refines() never reaching a non-semantic type. But Refines() has two exits, not one — its loop condition is

while (current is Semantic semantic && semantic.Declaration is SchemaSemanticType declaration)

so it stops both at a type it has already seen (a real cycle) and at a link whose Declaration is null (a name the schema does not declare). Both chains look identical from the call site.

So A refines B, B refines Missing, and schema.Validate() answered:

[Error] A: Semantic type 'A' refines itself, directly or through a cycle, so it is represented as nothing.
[Error] B: Semantic type refines 'Missing', which this schema does not declare.

There is no cycle anywhere. The first message sends the author looking for one, and it names A — which is fine — rather than B, which is the declaration with the typo in it.

The fix

SchemaSemanticType.RefinesItself() asks the question the message actually makes: did the chain come back to a type it had already seen?

public bool RefinesItself()
{
	SchemaSemanticType deepest = Refines().LastOrDefault() ?? this;
	return deepest.UnderlyingType is Semantic semantic && semantic.Declaration is not null;
}

The chain ended on a cycle exactly when the deepest type it reached still refines something that resolves, since that is the link Refines() refused to follow a second time. An unresolved tail leaves a Semantic with a null Declaration there instead, and is left to the declaration that names it — which keeps it one message rather than two, the same rule CLAUDE.md states for an unresolved semantic type generally.

It is written in terms of Refines() rather than walking again, and reuses the LastOrDefault() ?? this idiom already in Representation() so the direct A refines A case (which yields nothing) is covered by the same line.

Tests

• AnUnresolvedRefinementDeeperInAChainIsNotACycle — the two-link Weight → ForceMagnitude → Missing chain the issue describes, asserting exactly one issue, that it is the "does not declare" one, and that it is reported against ForceMagnitude. The existing AnUnresolvedRefinementIsRejected is depth 1 and never reaches this branch.
• ARefinementChainThatRevisitsATypeStillReportsItself — A → B → C → B, where the cycle does not include the type being asked, so the fix cannot narrow the check to self-reference.

Verified the first test fails without the change (Sequence contains more than one element — the spurious cycle error) and passes with it.

Schema.Test is 478/478 green and Schema.Cpp.Test is 102/102 green on net10.0. The net8.0/net9.0 legs could not be launched in this container (only the .NET 10 runtime is installed), so CI covers those.

🤖 Generated with Claude Code

https://claude.ai/code/session_01CCVJEzptkDVM2wWtGTY7Av

---

Generated by Claude Code

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
100.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud