Re: [Cp-tools-discuss] shape of the CVS

[Julian Scheid]

Nic Ferrier wrote:

> It works by generating a filelist of the source files that need to be
> compiled (which by default is done by comparing the .class file for
> the equivalent .java file) and then compiling that list with:
>
>   $(JAVAC) -d classes/ @filelist
>
> It's perfectly adequate for user builds and for developers.

Sounds OK. Would you care about setting up some configure.in and
Makefile.in appropriately? Alex do you agree?

> I am (just now in fact) working on some autoconf macros for java
> support, stuff like an AC_PROG_JAVAC and so forth. The resulting
> macros will be usable to configure the makefile.
>
> Who knows? My java rules might eventually make it into automake, but
> for now the Makefile is easily maintained separately.

Cool! I've been thinking of that too. Are you also looking to support
all different compilers and VMs? Have you seen Stephanos work?

> You can also add a trick that I do with paperclips. One of the
> problems when writing tools is that running the tool from java
> (or using command line tool support) requires the fully qualified
> name. With Paperclips I have a class called:
>
>    gnupaperclips
>
> which simply calls my normal "main" class. Note this class is in the
> top-level but very unlikely to cause any name clash.

My preferred approach is to have a conveniently named shell script
which does something like
#!/bin/sh
/path/to/java very.long.package.name.MainClass $*

For example, this makes it possible to invoke Gjdoc similar to
Javadoc, without having to say "java" first.

Julian