Instead they are retrieved
dynamically, with a pinch of caching during runtime. We already had some
fixes to make sure that compile time constanst were not inlined, but we
need to take into account that a lot of the Integer-based enums have
been replaced with actual enums.
Also, certain TreeMaker methods needed to be invoked dynamically with
reflection.
This needs to be reviewed, and if it turns out that these changes are too
dramatic, we should fork out a larger part of our code for specific JVM
versions.
After this was finalized I realize that we might consider moving entirely
to String based options, instead of inferring the options from the
provided enum. This setup does have the benefit of throwing exceptions when
options are not present.
Only Oracle provides builds for JDK8 atm, and there is no OpenJDK
build, so the setup is slightly different. Files have to be copied
manually from an installed JDK. For comparison purposes, an Oracle JDK7
option has been added too. This should be removed before merging to master.
a few or no withers at all, and adding them is a lot simpler than
removing them. This is a breaking change, but then that's why @Value
was in experimental in the first place.
* re-introduction of onMethod/onConstructor/onParam
* tests checking error/warnings rewritten to be more heuristic, in order to accomodate difference in messaging between java6 and java 7
* Ability to eliminate java's own output of erroneous error messages (heh); i.e. those messages that are invalidated by lombok's actions. This mechanism is used for onMethod/onConstructor/onParam
* First steps to unifying a billion setGeneratedBy calls into a single visitor traversal for eclipse' HandleGetter/Setter/Constructor/Wither
* To simplify 'zooming in' the tests on just a few files, added an 'accept' mechanism.
* Updated copyright headers of website to 2013.
* Removed most documentation and instead put in a link to the much more up to date and extensive documentation at http://projectlombok.org/features
* Getting ahead of ourselves a little, added notes on onConstructor/onParam/onMethod which we are about to add in the next few commits.
* Updated copyrights to 2012.
I fixed it by still compiling against a given javac (we can only ship one lombok.jar after all), but having the test task run with a given bootclasspath and a given javac.jar.
There are 2 tasks that download both rt.jar and javac.jar for either OpenJDK6 or OpenJDK7, and it writes a properties file with those locations. The test task will use this property file, and explain what you need to do if it is not there.
Incidentally, this brought to light issue 422: Delombok in java7 produces VerifyErrors.
Robbert Jan Grootjans