> -----Original Message-----
> From: Sam Ruby [mailto:[EMAIL PROTECTED]]
> Sent: Thursday, January 04, 2001 6:47 AM
> To: [EMAIL PROTECTED]
> Subject: Re: build.classpath and tinderbox builds
>
>
> Stefan Bodewig wrote:
> >
> > I think what Diane wanted to point you at was some
> > discussion we had a while back (and I think Jay
> > implemented the outcome, but I haven't looked at
> > his code).
>
> I took a look at the refactored Javac. The Jikes plugin was
> incomplete,
> and I commented on that here. Had that work been complete, I
> would have
> gladly folded it in.
I'm a little confused by this comment: "the jikes plugin was incomplete ... I've commented on that here". The only comment that you've made in the past about the refactoring changes is the following:
"P.S. After a quick glance, it appears that the Jikes code is missing the
output parsing and batching of input files to work around Windows
limitations."
I believe that this statement is incorrect. If you look at the Jikes plugin, it uses the executeExternalCompile method in DefaultCompilerAdapter which Jikes subclasses. Inside that method, if the command line is greater than 4096, then the files to be compiled are batched out to file.
Am I correct in believing that your only concern with the refactoring changes is the issue of batch-compilation for jikes? Has this info addressed your concerns with the refactoring?
>
> > We wanted to have two new attributes, something like
> > addruntime and addsystemclasspath. Looking at the
> > code you've committed it seems one could achieve the
> > desired effect with addsystemclasspath="false". Say
> > you have
>
> Not exactly, I really want the ability to control the order
> *outside* of
> the build.xml file.
Agreed, the refactoring does not address the order of items in the classpath. It only allows the user to remove the ant libraries and java system libraries from the classpath.
I'm a little confused as to how the position of the system libraries alters the results of your build. Can you provide a small example of how this is the case? I've looked at pages you've referred to previously (http://oss.software.ibm.com/developerworks/opensource/jakarta/proto/index.html) but because your builds don't provide the verbosity needed, this doesn't answer my confusion.
Just out of curiosity, are you using the bootclasspath attribute? I know of a bug in the current implementations of <javac> that if you provide something for the bootclasspath attribute, the system classes are still added to the classpath. This is something I hope to address after my refactoring is adopted (committers, when will it be accepted?) as this behaviour is different for each compiler.
Jay
