I haven't ever seriously used jde-complete b/c it didn't quite work
(Xemacs, linux) and I could be bothered to find out why. Well, I
finally looked closer and believe I've found the reason.
Symptoms:
The problem was that it wouldn't cycle through the possible completions:
instead it would emit a not-very-useful stub which I would then have
to edit. Also, it complained about an eldoc function not being defined.
so,
java.util.List l = new ArrayList();
l.i[M-x jde-complete-at-point] ===> l.indexOf(java.lang.Object : int);
&& a message about problems with eldoc.
Diagnosis:
The jde-complete.el source has this in it:
;;jde-eldoc for completion signatures
(eval-when-compile (require 'eldoc)
(require 'senator))
This makes it possible to compile jde-complete.el to jde-complete.elc.
BUT, there's no runtime (require 'eldoc) anywhere in the JDE. Since I
don't use eldoc (which is part of (x)emacs) in any of my other emacs
stuff, eldoc never gets loaded, and jde-complete's references to eldoc
code do not work. Hence, jde-complete does not work.
I don't know much about eval-when-compile, but I'm guessing that it
does get evaluated when you load jde-complete from the .el file
(as opposed to a byte-compiled .elc file). If this is true, it would
explain why this code might work for those who do not byte compile the
JDE, yet fail for those who do. [I'm speculating - I haven't looked into
this at all.]
At any rate, since semantic is a requirement for the JDE and since eldoc
is part of (x)emacs, I propose dropping the 'eval-when-compile' bit and
just having plain old
(require 'eldoc)
(require 'senator)
With this minor change, jde-complete works as advertised and has become
a very handy tool!
Eric
