testsuite problems on darwin was: [patch] HAVE_LC_MESSAGES f ornatSystem.cc

Boehm, Hans hans_boehm@hp.com
Mon Jan 21 10:12:00 GMT 2002


The read() wrapper is gone in my current tree anyway.  (I actually left it
inside a #if 0 as an example for anyone who wants to wrap system calls in
this way.)  It seemed to be causing more harm than good.  I believe the plan
for gcj is to no longer protect pointer-free pages when it can be avoided,
to enable incremental GC only if it can be avoided, and to avoid making
system calls from libgcj that write to the pointer-containing heap.  The
interface to support that is in the latest GC alpha releases.  In this was
gcj should never need the wrappers.

Hans

> -----Original Message-----
> From: Jeff Sturm [mailto:jsturm@one-point.com]
> Sent: Monday, January 21, 2002 9:52 AM
> To: Andreas Tobler
> Cc: Stan Shebs; tromey@redhat.com; GCC-Java
> Subject: Re: testsuite problems on darwin was: [patch] 
> HAVE_LC_MESSAGES
> fornatSystem.cc
> 
> 
> On Mon, 21 Jan 2002, Andreas Tobler wrote:
> > In other words, I have to get rid of the read() in 
> boehm-gc/os_dep.c for darwin?.
> > A first look into it showed me it's quite complicated to understand.
> 
> Undefining MPROTECT_VDB should do, provided you don't need or want
> incremental collection.
> 
> Jeff
> 



More information about the Java mailing list