idl2java

[marc Luce <ma...@blackbird.net>]

Is there an idl2java task out there?  The JDK ships with a reference
implementation of an idl2java compiler, right?  If there is not an idl2java
task already written, would it make sense for me to build one that is very
similar in design to the javac task?  I like the structure of that task and
the big difference would be rather than the set of supported compilers
(Modern, Classic, Jikes, etc) the idl2java compiler property would be a
class name (com.inprise.vbroker.Compiler for example).

How does this sound?  Any direction is appreciated.

Re: idl2java

[Stefan Bodewig <bo...@bost.de>]

marc Luce <ma...@blackbird.net> wrote:

> Is there an idl2java task out there?

There have been people working on it, I even vaguely remember code
being sent to ant-dev in May, maybe take a look at the archives.

> I like the structure of that task [javac]

but be prepared that magic properties will be replaced by something
better some time 8-)

> and the big difference would be rather than the set of supported
> compilers (Modern, Classic, Jikes, etc) the idl2java compiler
> property would be a class name (com.inprise.vbroker.Compiler for
> example).

Don't think this is true. Why would you want to restrict yourself to
idl2java compilers written in Java?

The main problem you face here - and I think this is why the effort
had been dropped back then - is that it is very difficult to maintain
a common set of options. 

idl2java compilers are rumored to be very different when it comes to
options (the "what is there" side of things, not the "how to
invoke"). 

If you really tried to write a generic idl2java task you'd have to
stick with the least common denominator (not too much probably) and
either allow for the user to specify additional arguments (a nested
<arg> element or similar) or introduce a bunch of magic properties
like we have in the jikes case.

Stefan