It's me again.
John, what you're describing is the behavior I wrote about on Wednesday. Discussion inline:
On 10/27/2017 03:03 AM, John O'Hara wrote:
Thank you for your responses. If I fill you in on a bit of background about what I have experienced, then it might help clarify the issue I see.
Background
I am running a JEE benchmark on EAP7.1 (using resteasy 3.0.24.Final). One of rest endpoints has a @Valid annotation on a method parameter for a complex object.
@POST
@Path("/foo")
@Consumes(MediaType.APPLICATIO
N_JSON )public void addFoo(@Valid Foo bar) throws Exception {
The Foo class contains different validation annotations, such as @NotNull, @Size, @Pattern and a custom validation annotation.
One test determines how WF/EAP handles invalid requests. When I invoke the endpoint above I see a non-deterministic response from the app server. If I invoke the endpoint with an invalid json object I receive a HTTP 400 exception as expected, but if I invoke the endpoint with a valid object, and *then* invoke the endpoint with an invalid object I receive a HTTP 500 error.
This exception only occurs on some property validation annotations and not all (e.g. @NotNull works as expected)
The test is being run in the context of CDI, and the HTTP 500 error is coming from the ejb3 subsystem. The ejb3 subsystem is wrapping the ConstraintValidationException raised by hibernate-validator.
Observations
In the call stack I capture [1], and WF thread dump [2] I can see;
The test is run in the context of CDI. The HTTP 500 response code message is caused by the ejb3 subsystem wrapping a ConstraintValidationException raised by hibernate-validator
The validator is not called in ResourceMethodInvoker.
invokeOnTarget() [3] as isValidatable is false [4], a ResteasyViolationException is not thrown
The request is validated again here [5], the validation exceptions are not acted upon in this method call and the reflected method (which happens to be the Foo class constructor) is called here [6]
Questions
Why is ResourceMethodInvoker.
isValidatable false in WF/EAP?
In the context of CDI, field and property injections are not carried out until a method is called on an object. I don't know why, but that's how it is. So it's possible to do parameter validation inline in MethodInjectorImpl.invoke(), but it's premature to do field/parameter validation in ResourceMethodInvoker.invokeOnTargetAfterFilter().
If isValidatable is false, and validation occurs again at MethodInjectorImpl.invoke() can we not check for validation errors and raise the correct exception there as well?
In effect, that's what is happening, but it's buried in org.jboss.resteasy.cdi.JaxrsInjectionTarget.inject(). In particular:
private void validate(HttpRequest request, T instance)
{
if (GetRestful.isRootResource(clazz))
{
ResteasyProviderFactory providerFactory = ResteasyProviderFactory.getInstance();
ContextResolver<GeneralValidatorCDI> resolver = providerFactory. getContextResolver( GeneralValidatorCDI.class, MediaType.WILDCARD_TYPE);
if (resolver != null)
{
validator = providerFactory.getContextResolver( GeneralValidatorCDI.class, MediaType.WILDCARD_TYPE). getContext(null); /// Hmmm, this looks redundant ...
}
if (validator != null && validator.isValidatableFromCDI(clazz))
{
validator.validate(request, instance);
validator.checkViolationsfromCDI( request);
}
}
}
Note that validator.checkViolationsfromCDI( request) will throw a ResteasyViolationException if any field/property/parameter violations have been detected.
-Ron
I have a standalone test that I will clean up and push out. There are some aspects of the test that can not be published publicly, so I need to modify them first.
Thanks,
John
[1] - see failing_4.stack
[2] - see WF_stackTrace.out
_______________________________________________ resteasy-dev mailing list resteasy-dev@lists.jboss.org https://lists.jboss.org/ mailman/listinfo/resteasy-dev
-- My company's smarter than your company (unless you work for Red Hat)
_______________________________________________
resteasy-dev mailing list
resteasy-dev@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/resteasy-dev
JOHN O'HARA
PRINCIPAL SOFTWARE ENGINEER
 |