use method handles for reflection

[cdietrich]

No description provided.

[merks]

This looks very nice and will perform better.

[szarnekow]

I think we can simplify this a bit more:

	private final static MethodHandle GET_ROOT_PATH_TO_RESOLVED_ENTRIES = findAccessor();
	
	private static MethodHandle findAccessor() {
		try {
			if (JavaCore.getPlugin().getBundle().getVersion().compareTo(new Version(3, 39, 0)) >= 0) {
				return MethodHandles.lookup().findGetter(PerProjectInfo.class, "rootPathToResolvedEntries", Map.class);
			} else {
				return MethodHandles.lookup().findVirtual(PerProjectInfo.class, "getRootPathToResolvedEntries", MethodType.methodType(Map.class));
		}
		} catch (ReflectiveOperationException e) {
			
			return MethodHandles.constant(Map.class, Collections.emptyMap());
		}
	}
	
	protected Map<IPath, IClasspathEntry> getRootPathToResolvedEntries(PerProjectInfo info) throws JavaModelException {
		try {
			return (Map<IPath, IClasspathEntry>) GET_ROOT_PATH_TO_RESOLVED_ENTRIES.invoke(info);
		} catch(Throwable t) {
			throw new JavaModelException(Status.error(t.getMessage(), t));
		}
	}

I'll push to this branch. WDYT?

[cdietrich]

@szarnekow do you think we still will be able to detect future breakages?
via AccessRestrictionInWorkspaceTest?

[cdietrich]

order was wrong. have reversed

[szarnekow]

@szarnekow do you think we still will be able to detect future breakages? via AccessRestrictionInWorkspaceTest?

We can remove the fallback to MethodHandles.constant and throw an exception. Or use Java assertions in the catch block (assuming the tests run with assertions enabled)

[szarnekow]

via AccessRestrictionInWorkspaceTest?

Yes, that'll certainly break if things change in an incompatible way.