Age | Commit message (Collapse) | Author |
|
improvement on some bit wrangling.
|
|
|
|
then eclipse would hang.
|
|
integration with patcher.
|
|
were in the path due to lack of decoding a URL.
|
|
class be invisible to most autocomplete dialogs and yet still work as an SPI provider; SPI lines are binary names (with dollars for inner classes), that's all. Name fixed and previous commit reverted.
|
|
class, doesn't work after all. Removed it and renamed the launching processor to _ShadowLaunchingAnnotationProcessor.
|
|
urls and back is an utter mystery' problems.
|
|
|
|
|
|
|
|
Right now the rules are:
* _IF_ a class is being loaded, sourced by a lombok-jar originating class, we FIRST search the lombok jar, and if we can’t find it, farm out the job to the originating equinox-side loader.
* _IF_ the equinox-side loader attempts to load a class, and it does NOT start with lombok, we don’t interfere and would never serve up any content from the lombok-jar (so if we have deps, they do NOT get loaded, by design). If it DOES start with lombok, we load it, and the loading class is SCL, not the equinox-side loader.
* getResource() to load classes did not work (because internally classes end in .SCL.lombok and not .class). This breaks a bunch of things. Fixed by having getResource() be aware that it should try rewriting any request for a .class to .SCL.lombok.
* launchified annotationprocessor, and cleaned up the launchified agent, which now, like all other launchers, just sets up classloader stuff and then calls into the lombok loader side to finish the actual processing, instead of trying to do it itself in a handicapped environment that can’t load much.
|
|
have to solve the problem with adding our shadow loader to the equinox infrastructure (solved in lombok currently by adding all of lombok to the bootclasspath), and all the public API still has to be kept as actual class files by build.xml. Currently it is all shadowed away.
|