Migrated from Google Code (issue 43)
👤 voidstar 🕗 Sep 02, 2009 at 01:35 UTC
What steps will reproduce the problem?
Please provide any additional information below.
In my case I'm using an older version of hibernate validator with it's
@ org.hibernate.validator.NotNull annotation. The newest version uses @ javax.validation.NotNull
which from what I can tell will have the same issue.
See this thread:
👤 reinierz 🕗 Sep 02, 2009 at 02:04 UTC
Well, it'll only choke on that particular variant of @ NotNull (javax.validation.NotNull). But, show-stopper. We'll fix
👤 r.spilker 🕗 Sep 02, 2009 at 08:27 UTC
To solve it right, we need to implement the bindings so we can tell what exact
NotNull is being used and maybe even check where it can be applied.
For now, I'm considering to NOT copy the NotNull annotation to the parameter because,
for as far as I know, the only other NotNull annotation in the wild is the one from
IDEA, and currently we don't support IntelliJ anyway. And not having the annotation
copied won't break anything.
👤 reinierz 🕗 Sep 02, 2009 at 08:31 UTC
Yup. For safety's and consistency's sake, we should probably stop copying the annotations everywhere (so, also
for the getter), and not just for the parameters of the setter and the constructor.
👤 jacek99 🕗 Sep 02, 2009 at 08:59 UTC
Wouldn't it be better to have some generic logic that would look at how an annotation
is defined and where it can be applied and use that in a common fashion to decide if
annotations should be copied around?
The same issue may pop up on other annotations that Lombok may have to deal with in
👤 reinierz 🕗 Sep 02, 2009 at 09:18 UTC
jacek99: Well, what kind of 'generic' logic are you referring to? If you're referring to: Check the target
ElementType and copy only if its legal - well, lombok can't do that - no type info. That's what Roel meant in
comment ﹟2 with "we need to implement the bindings so we can tell". It's planned - the suggested 'let's not
copy anything' is a short-term workaround.
👤 reinierz 🕗 Sep 02, 2009 at 23:41 UTC
Fixed in c46bf85 which will be rolled out in v0.8.5.
The "fix" simply reduces the feature to scan only for annotations named 'nonnull'. All annotations named nonnull
work the way we want them to (e.g. legal on parameters). The only commonly used notnull that ISNT a variant of
javax.validation.NotNull is IDEA's, but lombok doesn't support IDEA.
This is a workaround - once lombok supports type introspection we'll fix this properly.
👤 matt.deboer 🕗 Oct 17, 2011 at 19:48 UTC
This doesn't seem to be resolved (or has regressed?); using Lombok 0.10.1 with a maven 3.0.3 build, having a simple class with a single member which has @ javax.validation.constraints.NotNull applied causes compiler plugin to bomb with the aforementioned "annotation type not applicable to this kind of declaration".. this works correctly within Eclipse environment, but fails during Maven build.
(Compiler plugin is configured to use compilerVersion,source,target all = 1.6)
👤 reinierz 🕗 Oct 25, 2011 at 13:04 UTC
It's a regression introduced in 0.10.1. See issue #360 for more up to date discussion.
End of migration
added handling of @Builder annotation on the constructor/method and not on the class file