From: Peter Graves <peter@ar...> - 2005-07-04 13:55:24
ABCL 0.0.7 is now available:
> If all goes well, I'll try to put together a binary distribution for
Things did go well (in other respects), but at the last minute I ran
out of time, so 0.0.7 is still just a source release. The README file
in the root directory of the distribution contains build instructions.
Thanks to Kevin Reid for providing a starting point for Mac OS X build
support. I went on from there without benefit of a Mac, so in the end
it may or may not work. Please let me know.
The big new thing in this release is that in an emergency, ABCL can now
be used as a build host to build SBCL. If you have abcl in your PATH
and source for SBCL 0.9.2.12 (or later), you can just do
sh make.sh abcl
and when you come back from vacation, you should have a new SBCL. (The
build takes 1 hour 42 minutes on my AMD Athlon(tm) XP 2100+; I'll try
to make that go a bit quicker in future releases.) Thanks to Christophe
Rhodes for his advice and encouragement over the many months it took to
get the build-sbcl project to this point.
On non-Windows platforms, the abcl startup script that is generated by
the normal ABCL build now contains
to increase the maximum size of the Java heap. This (or something
similar) is necessary for building SBCL.
ABCL 0.0.7 fails 90 out of 21206 total tests in the GCL ANSI test
suite, compared to 77 out of 21085 tests for 0.0.6. The higher number
of failures doesn't reflect any significant regressions. The additional
failures come, for the most part, from new tests, plus a few failures
that are caused by the relaxation of initarg checking in 0.0.7's CLOS.
(The code that checked for bogus keyword arguments to MAKE-INSTANCE and
friends was broken in previous releases and sometimes interfered with
the execution of correct user code, so I've taken out the checking
until it can be fixed properly.)
Please report problems to the j development mailing list:
Thanks for your support.
From: Peter Graves <peter@ar...> - 2005-07-05 01:16:39
On Mon, 04 Jul 2005 at 10:48:23 -0500, Gerardo Horvilleur wrote:
> If you are setting the maximum heap size it might also be a good idea
> to set the initial heap size to the same size:
> java -Xmx256M -Xms256M
> If you know you will be using a large heap this avoids wasting a lot of
> CPU time doing a full garbage collection before each time the heap size
> is incremented.
For many purposes, you're not going to need anywhere near 256M of heap
(even ansi-tests is fine with 128M or less), and in those situations
it's probably better not to pre-allocate the full 256M (it might force
a machine with less memory into swap right away).
And in any case, it's not really a lot of CPU time that's wasted, maybe
a few seconds out of an sbcl build that takes 1 hour 42 minutes.
It probably does make sense to do something like -Xms8M, since the 2M
default isn't even enough for ABCL startup, and of course folks that
know they're going to use a lot of heap might want to do exactly as you