Heap fragmentation (Was: Debugging "Leaks" With Boehm-GC)

Wed Jan 18 07:26:00 GMT 2006

I can no longer find a reference to the NO_SIGSET macro. It looks to me
like it can disappear. I believe we always build with NO_SIGNALS
defined, which I believe means that these functions shouldn't be called
anyway.
I'll delete anything related to NO_SIGSET from the gc7alpha cvs tree.
(NO_SIGNALS is already gone there, since it became the only supported
option.)
If there is agreement that we should no longer do things differently for
cross-builds, I will change at least the gc7 tree correspondingly. This
was originally a compromise to try for a smaller configuration for
embedded builds. Since there wasn't a real option to control that, the
fact that you were doing a cross-build was deemed a hint that you were
targetting a small embedded environment. If we're adding explicit
options for smaller configurations, something like NO_DEBUGGING should
clearly be controlled by one of those, not by whether this is a
cross-build.
Hans
> -----Original Message-----
> From: java-owner@gcc.gnu.org [mailto:java-owner@gcc.gnu.org] 
> On Behalf Of Martin Egholm Nielsen
> Sent: Tuesday, January 17, 2006 1:08 PM
> To: java@gcc.gnu.org
> Subject: Re: Heap fragmentation (Was: Debugging "Leaks" With Boehm-GC)
>>> >>>You will need to build without NO_DEBUGGING, since that 
> also prevents 
> >>>GC_dump from being defined.
> > Martin> I hope it's enough to remove from the Makefile (and possibly
> > Martin> boehm-cflags) and do "make clean && make && make 
> install", and 
> > Martin> then redmake libgcj... That's what I'm doing now...
> > Hmm, does it appear in the Makefile? In the gcc tree at 
> least, it is 
> > set in configure.ac:
> Well, yes. In the Makefile in there is a nice 
> "-DNO_DEBUGGING=1" entry. 
> And it works just rebuilding "boehm-gc" and then "libjava" (and 
> installing)...
>> > if test -n "${with_cross_host}"; then
> > AC_DEFINE(NO_SIGSET,1,[use empty GC_disable_signals and 
> GC_enable_signals])
> > AC_DEFINE(NO_DEBUGGING,1,[removes GC_dump])
> > fi
> Oh, what is the "GC_disable_signals" and "GC_enable_signals" about?
>> BR,
> Martin
>>