ant-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Tom Cook <tc...@ardec.com.au>
Subject RE: WIBNI for Javac Task
Date Tue, 04 Jul 2000 00:23:04 GMT
Conor MacNeill writes:
 > > -----Original Message-----
 > > From: Tom Cook [mailto:tcook@ardec.com.au]
 > >
 > >
 > > I was under the impression that ant found each source file, and worked
 > > out where the destination file would end up, and then do comparisons
 > > on them. This works fairly faultlessly for me - are you using the 3.1
 > > release, or the latest CVS source? The release is hopelessly out
 > > of date...
 > >
 > > Regards
 > > Tom
 > 
 > The issue is that Erich's directory structure does not match his package
 > structure. I am not sure this is a good idea. When javac (the tool, not the
 > task) is compiling a class and it cannot find a class upon which the current
 > class depends, it attempts to find a source file, mapping the package
 > structure to the sourcepath. Have a look at "SEARCHING FOR TYPES" in the
 > Javac documentation. In all our projects, our source directory structure and
 > package structure match.

Forgive my persistence, and it's been a while since I hacked through
this code, but I thought ant decided which files to recompile _before_
passing them on to javac (at least for the classic compiler). I was
under the impression that it found all the source files specified,
worked out where they would end up (this bit may be a bit wonky if
your directory structure doesn't match, I guess) and compared
timestamps. Then if it needed recompiling, the file found it's way to
the javac command line. Isn't this the way it works? Or is my view of
the world going to be torn apart? ;-)

Regards
Tom

Mime
View raw message