Dominique Devienne wrote:
OK, to be more precise, it's dangerous to batch the files when one
explicitly resets the sourcepath in <javac>. You don't see a problem,
because SUN's Javac is "griddy", so if a source in batch#1 requires a
file in next batches, it will find it anyway through the sourcepath.
So batching might compile more files than you specify, because it will
follow the web of dependencies of the sources.
I got the same conclusion. Nevertheless, in my case with properly
chosen batch size I got results about twice as fast (~9-10 vs ~18-20
minutes for
full rebuild) and with 1/3 of memory (~0.5Gb vs 1.5+ Gb) which was needed
with one-step compilation.
So it's safer than I initialy thought, but might not do what you
thought it does either. In any case, using batching + sourcepath *is*
dangerous. --DD
Well, I agree that this is not safe in all cases but it does what it
should do,
at least in my case. Perhaps it is too early to submit it for public use
though.
Thanks a lot for valuable discussion.
Regards,
Sergey.
*---------------------------------------------
[EMAIL PROTECTED] http://es.os2.ru/
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
