On Fri, 20 Jul 2001, Paul Speed wrote:
>
>
> "Craig R. McClanahan" wrote:
> >
> > There are some pretty intense discussions going on about this, and the
> > story hasn't yet been finished ...
> >
> > On the other hand, the "new" compiler entry point has an absolutely
> > horrible feature (from the point of view of Jasper) -- you have to modify
> > System.out and System.err to redirect the compiler output. Unless system
> > variables are unique per classloader (they weren't last time I checked),
> > that means we can only do one compile at a time :-(.
>
> This concerns me for a lot of reasons. Mostly because it's a sign
> of a poorly architected program. A compiler that does not abstract its
> error reporting is very poorly designed. What happens when Java has
> a logging API and they want to use that instead?
>
> I'm one of the biggest Sun supporters you'll find. I think a lot of
> bright engineers have come up with some very cool stuff. However,
> this "feature" is one of the most boneheaded things I've ever heard. ;)
>
I think it's more an issue of not even thinking about needs beyond a
command line compiler (which is all it was for in the first place --
remember we're using an *unsupported* internal entry point). It's as much
our (Tomcat and Ant communities, as well as others who do internal
compilations) fault for not being more actively involved in making sure
the J2SE folks understand our needs as anything else.
> -Paul Speed
>
Craig
