So the issue is a more general one than I thought.
We'll run into the same situation, too, when there is an error in BV's
validation.xml or any constraint mapping file (e.g. a reference to a
non-existent MessageInterpolator). That error will currently be
suppressed with validation-mode auto.
I'm wondering whether in BV 2.0 we should define a new sub-type of
ValidationException: NoProviderFoundException which is raised by
javax.validation.Validation in case no BV provider is found.
Then JPA/ORM could specifically ignore this exception in
validation-mode auto but let all other exceptions bubble up, letting
the user know early on about any issues they have with their
validation provider or its config.
Thoughts?
2017-02-06 15:53 GMT+01:00 Emmanuel Bernard <emmanuel(a)hibernate.org>:
+1 to fail fast with the explicit error in that case.
> On 1 Feb 2017, at 11:33, Gunnar Morling <gunnar(a)hibernate.org> wrote:
>
> Hi,
>
> JPA defines for validation mode "auto" that bean validation must occur
> if a BV provider is present and that no validation shall occur
> otherwise.
>
> What should happen though if a BV provider such as HV is present but
> it fails to bootstrap? In case of HV this happens if no expression
> language implementation can be found.
>
> Currently, the user has a very hard time to find out about this, as
> this exception essentially is suppressed (for mode "callback" the
> exception is raised).
>
> Should we raise a specific exception in HV if it cannot be
> bootstrapped? In ORM, we could handle that one specifically and raise
> it also if for validation mode "auto" (would have to happen
> reflectively, though, as to avoid a hard dependency).
>
> I can do this change but first wanted to make sure that this is inline
> with what you all think should be done.
>
> Thanks,
>
> --Gunnar
> _______________________________________________
> hibernate-dev mailing list
> hibernate-dev(a)lists.jboss.org
>
https://lists.jboss.org/mailman/listinfo/hibernate-dev