| Age | Commit message (Collapse) | Author |
|---|
|
To support the new numeric literal syntax of Java 7, the constructor of IntLiteral got replaced by a factory-method. So we updated the affected handlers to work in both worlds.
|
|
|
|
skipPrintAst was a singleton global, so, yes, thread race issues all over the place.
|
|
print once only)
|
|
Looks like eclipse sometimes sets this flag even though
all method bodies are not in fact parsed yet, as a shortcut to try and force future eclipse parse runs to simply skip doing it
as a timesaver. D'oh.
|
|
"canEqual(java.lang.Object)" was not in the blacklist of HandleDelegate
(only javac, for eclipse it was already present)
|
|
Between javac 1.6 and 1.7 the type of the field JCClassDecl.extending has changed from JCTree to JCExpression, but the method getExtendsClause() still returns a JCTree. So we use this for now.
|
|
|
|
|
|
classpath, the 'right' BaseFileObject wrapper is now used.
|
|
|
|
this out in 0.10.0-RC2.
|
|
environment into doing multiple rounds and disabling certain reparses in netbeans. Thanks to Jan Lahoda for researching this issue and supplying us with a patch.
|
|
|
|
them.
|
|
|
|
240.
|
|
|
|
|
|
|
|
|
|
even if its a post-diet handler.
|
|
interfaces; there were too many 'option' methods (methods where you return either true or false), so we really needed the ability to pick a default.
|
|
implements Foo<T>. Fixed this.
|
|
164' escape to not do that in ecj was no longer relevant.
|
|
because the 'isFullParse' boolean was erronously set on a full rebuild, forced by HandleDelegate.
HandleDelegate has also been updated to not do so much work if there's no @Delegate in a source file.
|
|
especially class files built with java7.
Fixes issue #188
|
|
(java7)
|
|
Synchronized and SneakyThrows
|
|
|
|
annotation processing offline.
|
|
a severe error in PatchDelegate(issue #211) where classes that didn't use lombok at all could not be build due to a NullPointerException.
|
|
handled is only done if we are -actually- going to call a handler.
|
|
via the LombokAST object. Instead its tracked more directly in an attempt
to avoid having to write all handlers as idempotent, and just in case
issue #164 is a race condition (the handled-or-not is a synchronized CAS check).
This does break API for other plugins, but the fix is trivial: Just make your
'handle' method return void. That 'we won't call you again' business in the decks
never quite worked right anyway.
Also, you might want to call Javac.(recursive)setHandledBy when you generate nodes, now.
|
|
HandleDelegate was called multiple times for the same field resulting in an error saying the delegate method was already defined.
Also added a test for @Delegate that uncovered this issue.
|
|
it won't actually work right; method-level generics would break.
The new way is to use -javaagent:lombok.jar=ECJ in addition to -cp lombok.jar
|
|
against situations in which lombok can't possibly work, since there is not enough of eclipse on the classpath. Major example of this is when we are called through jsps in the eclipse help.
Also added an extra check before the patchval is performed by checking if previous patch calls failed in TransformEclipseAST.
This fixes Issue #207.
|
|
|
|
|
|
works.
|
|
preclude them from being generated, which means you get duplicate method errors.
excludes=Types has been added to counteract this. Once we figure out how to
resolve method sigs out of order we can go back to the original plan.
|
|
|
|
NoSuchMethodError: CastExpression.<init>
This commit fixes this (now lombok works both <3.7 and 3.7).
fixes issue #206
|
|
a user runs into issue #164, we can get a more useful stracktrace from them.
|
|
|
|
Lombok.preventNullAnalysis but go with Collections.singletonList(expr).get(0)
instead; while this does create a pointless object, it doesn't cause a clash
when eclipse has lombok 0.10 installed but the project uses 0.9, which doesn't
have preventNullAnalysis. Eventually, once 0.9 is long forgotten, this can be reverted.
|
|
|
|
JavacProcessingEnvironment would occur when using the m2eclipse plugin, for example when removing an entry from the build path.
The actual change is small but this took quite some searching. m2eclipse uses plexus, and plexus uses a custom classloader, which means lombok can't find the JavacProcessingEnvironment loaded by that classloader.
We fix it here by adding lombok to that custom classloader. Perhaps more die-hard m2eclipse users find a problem with this approach, but assuming these plexus compile runs are stand-alone, this should work great.
fixed by Roel and Reinier.
|
|
|
|
'isFoo', 'hasFoo', or 'getFoo' would trigger specialized handling for @Getter/@Setter. However, this special handling broke the bean spec, and has been simplified: Only fields named 'isFoo', and only if that field's type is 'boolean', results in both 'isFoo' and 'foo' being considered as possible property names for this property, with 'foo' preferred, so that @Getter boolean isFoo will generate setFoo and isFoo methods, not setIsFoo and isIsFoo.
Fixes issue #148
|
