BVTCK-188 Allow bootstrap exceptions to be thrown at deployment

[gsmet]

https://hibernate.atlassian.net/browse/BVTCK-188 https://hibernate.atlassian.net/browse/BVTCK-190 https://hibernate.atlassian.net/browse/BVTCK-191

[aguibert]

@gsmet here is the full set of classes I found that had this issue:

org.hibernate.beanvalidation.tck.tests.bootstrap.BootstrapConfigurationWithEmptyValidatedExecutableTypesTest
org.hibernate.beanvalidation.tck.tests.bootstrap.customprovider.BootstrapNonAvailableValidationProviderTest
org.hibernate.beanvalidation.tck.tests.valueextraction.declaration.MultipleValueExtractorsInValidationXmlForSameTypeAndTypeArgumentTest
org.hibernate.beanvalidation.tck.tests.valueextraction.declaration.ValueExtractorWithNoPublicNoArgConstructorInValidationXmlTest
org.hibernate.beanvalidation.tck.tests.xmlconfiguration.ClockProviderSpecifiedInValidationXmlNoDefaultConstructorTest
org.hibernate.beanvalidation.tck.tests.xmlconfiguration.constraintdeclaration.ConfiguredBeanNotInClassPathTest
org.hibernate.beanvalidation.tck.tests.xmlconfiguration.constraintdeclaration.fieldlevel.WrongFieldNameTest
org.hibernate.beanvalidation.tck.tests.xmlconfiguration.constraintdeclaration.MandatoryNameAttributeTest
org.hibernate.beanvalidation.tck.tests.xmlconfiguration.constraintdeclaration.MissingMandatoryElementTest
org.hibernate.beanvalidation.tck.tests.xmlconfiguration.constraintdeclaration.propertylevel.WrongPropertyNameTest
org.hibernate.beanvalidation.tck.tests.xmlconfiguration.constraintdeclaration.ReservedElementNameTest
org.hibernate.beanvalidation.tck.tests.xmlconfiguration.constructorvalidation.UnknownConstructorValidationTest
org.hibernate.beanvalidation.tck.tests.xmlconfiguration.invalid.InvalidValidationXmlTest
org.hibernate.beanvalidation.tck.tests.xmlconfiguration.MessageInterpolatorSpecifiedInValidationXmlNoDefaultConstructorTest
org.hibernate.beanvalidation.tck.tests.xmlconfiguration.methodvalidation.MethodAsGetterAndMethodNodeTest
org.hibernate.beanvalidation.tck.tests.xmlconfiguration.methodvalidation.UnknownMethodValidationTest
org.hibernate.beanvalidation.tck.tests.xmlconfiguration.MissingClassNameOnBeanNodeTest
org.hibernate.beanvalidation.tck.tests.xmlconfiguration.TraversableResolverSpecifiedInValidationXmlNoDefaultConstructorTest
org.hibernate.beanvalidation.tck.tests.xmlconfiguration.versioning.UnknownVersionInValidationXmlTest

Also, I noticed a few other tests were failing for me because the test apps are missing some required classes:

// these apps need to include org.hibernate.beanvalidation.tck.tests.xmlconfiguration.versioning.DummyMessageInterpolator
org.hibernate.beanvalidation.tck.tests.xmlconfiguration.versioning.Version10InValidationXmlTest
org.hibernate.beanvalidation.tck.tests.xmlconfiguration.versioning.Version11InValidationXmlTest

// this app needs to include the inner class CustomMessageInterpolatorTest.DummyMessageInterpolator
org.hibernate.beanvalidation.tck.tests.bootstrap.CustomMessageInterpolatorTest

// this app needs to include org.hibernate.beanvalidation.tck.tests.xmlconfiguration.versioning.DummyClockProvider
org.hibernate.beanvalidation.tck.tests.xmlconfiguration.versioning.Version20InValidationXmlTest

[aguibert]

I pulled this change in locally and it seems to be working. I need to make a few additional tweaks on top of it, which I can contribute via another PR.

[gsmet]

@aguibert please wait for a bit. I'm working on it and will push a squashed version soon. Will let you know.

[gsmet]

I already updated all the tests you mentioned. I'm checking the issue you reported with the missing classes (I'm a bit surprised because we are running the TCK tests in WildFly with proper isolation but will check anyway).

[gsmet]

@aguibert so indeed, you're right on the XML tests but I have a hard time understanding the issue with this one:

// this app needs to include the inner class CustomMessageInterpolatorTest.DummyMessageInterpolator
org.hibernate.beanvalidation.tck.tests.bootstrap.CustomMessageInterpolatorTest

As we include the test, the inner class should be copied with the test.

[gsmet]

@aguibert I just force-pushed a new version with all the tests you mentioned fixed (and tested everything in WF container mode - so I consider this version OK for us).

Would appreciate if you could validate it works OK for you too.

The only issue I haven't fixed so far is this one:

// this app needs to include the inner class CustomMessageInterpolatorTest.DummyMessageInterpolator
org.hibernate.beanvalidation.tck.tests.bootstrap.CustomMessageInterpolatorTest

as I don't understand why it wouldn't work as is.

I'm finished with my changes for now so feel free to contribute additional work on top of it.

[aguibert]

@gsmet you're correct, I double checked CustomMessageInterpolatorTest and no change is needed. I had thought initially it wouldn't package the DummyMessageInterpolator and Person inner classes, but it does.

[gsmet]

@aguibert so do you pass the TCK with the version in this PR, now? Are other changes required?

Thanks for your feedback!

[aguibert]

@gsmet All of the TCK-side issues appear to be resolved now. Thanks! There was one minor tweak I had to make so the TCK builds on Windows, but I can contribute that separately.

I've got OpenLiberty almost passing the TCK now with your most recent set of changes from this PR. Currently 1036 tests pass. Still need to get the JavaFX stuff sorted out, and then there is one additional issue with OpenLiberty where specifying an app-defined <default-provider> in validation.xml does not work, but that'll need to be a fix I need to make in the OpenLiberty code.

[gsmet]

@aguibert I integrated a trimmed down version of your patch for the Windows support in this PR.

Could you please validate that everything is OK for you with this branch?

Thanks!

[gsmet]

@aguibert

I added a few commits to be on the safe side. I'm finished on this.

Here is what I suggest:

• except if you really need a blessed version, we wait for you to have a full successful run of the TCK;
• once you have sorted out the JavaFX issue and the issue in OpenLiberty, you give us your go and we release a new TCK.

If you really need a new version, we could release earlier but I really think it would be better to be sure all your issues are sorted out before making a new release.

Feedback welcome!

[aguibert]

hi @gsmet, it will probably be at least a week before I have time to work through the remaining TCK issues with OpenLiberty. Until then I can work off of a local snapshot with these changes

[gsmet]

If it works for you to work with the snapshot, let's do that.

And we'll release once we are totally sure we fixed all your issues.

[aguibert]

hi @gsmet just wanted to confirm that this PR plus the exclusion of the JavaFX tests gets OpenLiberty 100% passing for the BeanVal 2.0 TCK. Thanks for all of your help on this!

[gsmet]

@aguibert by JavaFX tests exclusion, you mean the other PR? Would be nice if you could just get the branch from the other PR, build it and test the TCK (just checking, it's what you're done so that we are sure everything is OK before releasing 2.0.2.Final).