Re: Kawa 1.13 compilation problem, possible regression

[Per Bothner <per at bothner dot com>]

On 02/04/2013 01:23 PM, Per Bothner wrote:
> > I was not able to compile the scheme part of the Kawa project because
> > the compiler is getting out of Heap Space when trying to compile the
> > file `gui.scm`.
> > ....
> > (compiling XML.scm to gnu.kawa.slib.XML)
> > (compiling XStrings.scm to gnu.kawa.slib.XStrings)
> > (compiling gui.scm to gnu.kawa.slib.gui)
> > gui.scm:152: internal error while compiling gui.scm
> > java.lang.OutOfMemoryError: Java heap space
> >          at java.util.Arrays.copyOf(Arrays.java:2760)
> >          at java.util.Arrays.copyOf(Arrays.java:2734)
> >          at java.util.ArrayList.ensureCapacity(ArrayList.java:167)
> >          at java.util.ArrayList.addAll(ArrayList.java:474)
> >          at gnu.expr.PushApply.forkNext(PushApply.java:276)
> >          at gnu.expr.PushApply.forkPop(PushApply.java:284)
> >          at gnu.expr.PushApply.visitIfExp(PushApply.java:138)
> >          at gnu.expr.PushApply.visitIfExp(PushApply.java:23)
>
> This is something I've seen but I haven't quite been able to
> pin it down: Sometimes it seems a problem, but when I try to
> narrow it down to when it happens, then it disappears.
>
> However, I have a general idea how this can be fixed.  The idea
> is to replace this data-structure:
>    ArrayList<HashSet<LambdaExp>> canFinishCondition
> by a Map<LambdaExp,IntNum>, where the IntNum result is a
> bitmap of "paths that can finish".  I have reason to believe
> this would behave better, but the devil is in the details.

I implemented this "inversion" of the data structures, and
it seems to solve the problem: Kawa doesn't run out of
heap space when setting a low value for -Xmx.
--
	--Per Bothner
per@bothner.com   http://per.bothner.com/