Memory problem

[Boehm, Hans]

There's a slight generalization of this in the GC7.0alpha tree now.  See

http://cvs.sourceforge.net/viewcvs.py/bdwgc/bdwgc/pthread_support.c?rev=
1.1.1.1&view=auto

I'm not sure that's enough to make everyone happy, though.  I think
there were other requests to run Java code in threads that predated the
libgcj startup.  That really requires a thread registration facility in
the GC.  That's also in GC7. (See GC_register_my_thread in
http://cvs.sourceforge.net/viewcvs.py/bdwgc/bdwgc/include/gc.h?rev=1.2&v
iew=auto .)

The problem is that that tree doesn't really have enough miles on it.
And I'm sure it currently breaks on a bunch of platforms.  And the
import is likely to be a mess, since a large fraction of the source
lines changed.  (It finally drops K&R compatibility among other major
changes.)

On the other hand, if something is pressing, it might be reasonable to
backport pieces.

Hans

> -----Original Message-----
> From: Anthony Green [mailto:green@redhat.com] 
> Sent: Friday, December 09, 2005 4:12 AM
> To: Boehm, Hans
> Cc: Gabriel Bianco; java@gcc.gnu.org
> Subject: RE: Memory problem
> 
> 
> On Thu, 2005-12-08 at 15:08 -0800, Boehm, Hans wrote:
> > That's a standing issue.  Currently it's not recommended that you 
> > create a thread from C++ code and then run Java in it.  At least if 
> > the thread creation is done in a library outside your 
> control, so that 
> > it can't be turned into a GC_pthread_create.  We're slowly moving 
> > towards a solution.
> 
> We need this for RSSOwl, Eclipse and perhaps more.
> 
> I proposed a solution here: 
> https://www.redhat.com/archives/fedora-devel-java-list/2005-Ju
ly/msg00009.html

Is this the kind of solution you're looking at?

Thanks,

AG