Treat only dots outside {variables} as childpath separators

[daedeloth]

ResourceDefinitionValidator skips dotted field names, because those set the child on some other entity whose class is not knowable from the definition.

It tested the raw field name — but reply:{context.model} contains a dot inside its variable, so every parameterised relationship was skipped before baseName() ever ran. The regression test shipped with the 1.11.1 fix used child:{context.model}, so it passed for the wrong reason: the skip caught it, not the code under test.

Dots inside {...} are now stripped before the check. nested.child is still skipped; reply:{context.model} is now actually checked.

Impact is false negatives only — those fields went unchecked rather than being wrongly reported, so no valid definition was ever flagged.

Verified load-bearing: restoring the raw-string check fails the new test. Suite green at 68 tests.

Found by the coverage work in #28, which independently spotted that baseName() was unreachable for the shape it was written for.

[daedeloth]

ResourceDefinitionValidator skips dotted field names, because those set the child on some other entity whose class is not knowable from the definition.

It tested the raw field name — but reply:{context.model} contains a dot inside its variable, so every parameterised relationship was skipped before baseName() ever ran. The regression test shipped with the 1.11.1 fix used child:{context.model}, so it passed for the wrong reason: the skip caught it, not the code under test.

Dots inside {...} are now stripped before the check. nested.child is still skipped; reply:{context.model} is now actually checked.

Impact is false negatives only — those fields went unchecked rather than being wrongly reported, so no valid definition was ever flagged.

Verified load-bearing: restoring the raw-string check fails the new test. Suite green at 68 tests.

Found by the coverage work in #28, which independently spotted that baseName() was unreachable for the shape it was written for.

[daedeloth]

ResourceDefinitionValidator skips dotted field names, because those set the child on some other entity whose class is not knowable from the definition.

It tested the raw field name — but reply:{context.model} contains a dot inside its variable, so every parameterised relationship was skipped before baseName() ever ran. The regression test shipped with the 1.11.1 fix used child:{context.model}, so it passed for the wrong reason: the skip caught it, not the code under test.

Dots inside {...} are now stripped before the check. nested.child is still skipped; reply:{context.model} is now actually checked.

Impact is false negatives only — those fields went unchecked rather than being wrongly reported, so no valid definition was ever flagged.

Verified load-bearing: restoring the raw-string check fails the new test. Suite green at 68 tests.

Found by the coverage work in #28, which independently spotted that baseName() was unreachable for the shape it was written for.

[daedeloth]

ResourceDefinitionValidator skips dotted field names, because those set the child on some other entity whose class is not knowable from the definition.

It tested the raw field name — but reply:{context.model} contains a dot inside its variable, so every parameterised relationship was skipped before baseName() ever ran. The regression test shipped with the 1.11.1 fix used child:{context.model}, so it passed for the wrong reason: the skip caught it, not the code under test.

Dots inside {...} are now stripped before the check. nested.child is still skipped; reply:{context.model} is now actually checked.

Impact is false negatives only — those fields went unchecked rather than being wrongly reported, so no valid definition was ever flagged.

Verified load-bearing: restoring the raw-string check fails the new test. Suite green at 68 tests.

Found by the coverage work in #28, which independently spotted that baseName() was unreachable for the shape it was written for.

[daedeloth]

ResourceDefinitionValidator skips dotted field names, because those set the child on some other entity whose class is not knowable from the definition.

It tested the raw field name — but reply:{context.model} contains a dot inside its variable, so every parameterised relationship was skipped before baseName() ever ran. The regression test shipped with the 1.11.1 fix used child:{context.model}, so it passed for the wrong reason: the skip caught it, not the code under test.

Dots inside {...} are now stripped before the check. nested.child is still skipped; reply:{context.model} is now actually checked.

Impact is false negatives only — those fields went unchecked rather than being wrongly reported, so no valid definition was ever flagged.

Verified load-bearing: restoring the raw-string check fails the new test. Suite green at 68 tests.

Found by the coverage work in #28, which independently spotted that baseName() was unreachable for the shape it was written for.

[daedeloth]

ResourceDefinitionValidator skips dotted field names, because those set the child on some other entity whose class is not knowable from the definition.

It tested the raw field name — but reply:{context.model} contains a dot inside its variable, so every parameterised relationship was skipped before baseName() ever ran. The regression test shipped with the 1.11.1 fix used child:{context.model}, so it passed for the wrong reason: the skip caught it, not the code under test.

Dots inside {...} are now stripped before the check. nested.child is still skipped; reply:{context.model} is now actually checked.

Impact is false negatives only — those fields went unchecked rather than being wrongly reported, so no valid definition was ever flagged.

Verified load-bearing: restoring the raw-string check fails the new test. Suite green at 68 tests.

Found by the coverage work in #28, which independently spotted that baseName() was unreachable for the shape it was written for.

[daedeloth]

ResourceDefinitionValidator skips dotted field names, because those set the child on some other entity whose class is not knowable from the definition.

It tested the raw field name — but reply:{context.model} contains a dot inside its variable, so every parameterised relationship was skipped before baseName() ever ran. The regression test shipped with the 1.11.1 fix used child:{context.model}, so it passed for the wrong reason: the skip caught it, not the code under test.

Dots inside {...} are now stripped before the check. nested.child is still skipped; reply:{context.model} is now actually checked.

Impact is false negatives only — those fields went unchecked rather than being wrongly reported, so no valid definition was ever flagged.

Verified load-bearing: restoring the raw-string check fails the new test. Suite green at 68 tests.

Found by the coverage work in #28, which independently spotted that baseName() was unreachable for the shape it was written for.

[daedeloth]

ResourceDefinitionValidator skips dotted field names, because those set the child on some other entity whose class is not knowable from the definition.

It tested the raw field name — but reply:{context.model} contains a dot inside its variable, so every parameterised relationship was skipped before baseName() ever ran. The regression test shipped with the 1.11.1 fix used child:{context.model}, so it passed for the wrong reason: the skip caught it, not the code under test.

Dots inside {...} are now stripped before the check. nested.child is still skipped; reply:{context.model} is now actually checked.

Impact is false negatives only — those fields went unchecked rather than being wrongly reported, so no valid definition was ever flagged.

Verified load-bearing: restoring the raw-string check fails the new test. Suite green at 68 tests.

Found by the coverage work in #28, which independently spotted that baseName() was unreachable for the shape it was written for.