UnexpectedTypeException: Cannot find a constraint validator for constraint

[mancze]

Expected Behavior

Validator should not throw the exception and behave uniformly on subsequent calls.

Actual Behaviour

Exception is thrown when validator.validate(validBean) is being validated:

jakarta.validation.UnexpectedTypeException: Cannot find a constraint validator for constraint:

Steps To Reproduce

Included in the example application including naive workaround. Run tests to identify bug.

1. Use Micronaut validation + Jakarta validation annotations
2. Declare a bean with validated properties
3. Inject Validator bean into your tests
4. Call validator.validate() with your valid bean
5. Exception is thrown
6. Catch it and re-run validator.validate()
7. No exception is thrown

Steps 5 and 7 produce inconsistent validation behaviour.

I think that possible cause is this line:

micronaut-validation/validation/src/main/java/io/micronaut/validation/validator/constraints/DefaultConstraintValidators.java

Line 151 in 89efd63

validatorCache.put(key, ConstraintValidator.VALID);

A ConstraintValidator.VALID is put to the validator's cache but Optional.empty() is returned. Subsequent calls of the method return the cached ConstraintValidator.VALID which is the reason of the inconsistent behaviour.

Environment Information

No response

Example Application

https://github.com/mancze/micronaut-sandbox/tree/repro/constraint-validation-not-found

Version

4.3.2

[dstepanov]

The error is correct. You don't have a validator defined. Remove @Constraint if you want just to apply other constraints.

[mancze]

Thanks. Removal of @Constraint(validatedBy = { }) seems to have helped.

We were using it because it was documented so in docs for older version and it seems it changed in v4.1.0:
3cee60e#diff-70e97b012bcbcfdcc57ed2c42af87338142c473a8200234add37f3920f2557ca

I didn't notice this change during migration from Microanut 3 to a recent version (4.3.2) of Micronaut 4. Does it mean that @Constraint shouldn't be applied if validator of the bean comes from the factory method?

[mancze]

Also, I observe some inconsistency related to the validation cache. With @Constraint(validatedBy = { }) annotation preserved and the test case being modified to:

try {
	assertThat(validator.validate(bean)).isEmpty();
} catch (Exception e) {
	// retry
	assertThat(validator.validate(bean)).isEmpty();
}

(mancze/micronaut-sandbox@e311721)

then only the first validation fails and the second one in the catch block not. The behaviour is not consistent.