This is the mail archive of the
java-patches@gcc.gnu.org
mailing list for the Java project.
RE: Merge GC 6.3
- From: "Boehm, Hans" <hans dot boehm at hp dot com>
- To: "'Bryce McKinlay'" <mckinlay at redhat dot com>
- Cc: java-patches at gcc dot gnu dot org
- Date: Fri, 13 Aug 2004 17:34:19 -0700
- Subject: RE: Merge GC 6.3
Thanks.
Some quick comments on remaining differences:
Mips_sgi_mach_dep.s: I think I remember changing this in the upstream sources. Based on an earlier discussion, the upstream sources are correct, assuming <sgidefs.h> or its equivalent is transitively included. If not, I think both are broken. It would be good if someone with access to an Irix machine could confirm, but based on what I know, I'd currently vote for the upstream version.
Pthread_support.c: I think pthread_support.h includes gc_priv.h, which includes gc.h. The difference here really doesn't matter, and could be removed.
Gc.h: I think the GC_init declaration is redundant. It's already there in a different place in the upstream source. (IIRC this used to be a bug in the upstream source.)
Gc_locks.h: This is probably a bit messed up, since someone updated the GC source to the gcc intriniscs, and the Intel compiler support ended up getting dropped in the process. I think both versions are correct in their context, so I'd be inclined to ignore it for now. Gc7.0 will use the atomic_ops package from www.hpl.hp.com/research/linux/qprof. (I'm hoping to eventually talk you guys into doing the same thing for the rest of libgcj.) Thus this is all slated to disappear.
gcconfig.h: I absorbed most of the changes in the upstream source, though I may have left some white space issues for next time.
Hans
> -----Original Message-----
> From: java-patches-owner@gcc.gnu.org
> [mailto:java-patches-owner@gcc.gnu.org]On Behalf Of Bryce McKinlay
> Sent: Friday, August 13, 2004 3:59 PM
> To: java-patches@gcc.gnu.org
> Subject: FYI: Merge GC 6.3
>
>
> I'm committing a merge of the Boehm GC 6.3 distribution. The
> patch below
> show the current diff between the GC in the gcc tree and the upstream
> sources, except for the configure/Makefile stuff.
> There was an autoconf issue that popped up - autoconf was
> putting some
> internal define's, such as PACKAGE_NAME, into the boehm-cflags file
> which is read by libgcj's configure. This resulted in duplicate
> definition warnings because those defines conflicted with the ones in
> libgcj's own config.h. The solution was to have the boehm-gc
> configure
> generate a gc_config.h file containing the definitions, and
> include that
> from gc.h. This required a few minor tweaks to other GC
> files, as follows:
> 2004-08-13 Bryce McKinlay <mckinlay@redhat.com>
> * gcj_mlc.c: Check #ifdef GC_GCJ_SUPPORT after including headers.
> * specific.c: Check #ifdef GC_LINUX_THREADS after
> including headers.
> * include/gc_config_macros.h: Remove backward-compatibility
> redefinitions of GC_ names.
> * include/gc.h: Include <gc_config.h>.
>