Re: [Gc] GC 7.2alpha2

Ivan Maidanski ivmai at mail.ru
Sat Jun 13 01:40:43 PDT 2009


Hi!

"Boehm, Hans" <hans.boehm at hp.com> wrote:
> I finally put another snapshot of the GC in the usual place (http://www.hpl.hp.com/personal/Hans_Boehm/gc/gc_source/)

Missing files in the tar ball:
- tests/trace_test.c,
- Makefile,
- autogen.sh,
- configure_atomic_ops.sh.

Things that shouldn't go to tar (IMHO):
- "CVS" folders (in libatomic_ops-1.2);
- libatomic_ops-1.2/autom4te.cache.
- libatomic_ops-1.2/src/atomic_ops.h.orig (duplicates atomic_ops.h),
- libatomic_ops-1.2/src/atomic_ops/generalize.h.orig (minor changes),
- libatomic_ops-1.2/src/atomic_ops/sysdeps/ao_t_is_int.h.orig (minor),
- libatomic_ops-1.2/src/atomic_ops/sysdeps/Makefile.am.orig,
- libatomic_ops-1.2/src/atomic_ops/sysdeps/gcc/x86.h.orig (minor).

Also look for powerpc.h.orig.

Missing file in CVS:
- bdw-gc.pc

>
> I only tested on 3 or 4 platforms, but it appears to be in decent shape.  The major outstanding problem I'm aware of are the occasional failures with GetWriteWatch incremental GC under Windows.  I'm still not quite caught up with patches, so there may be others.

Please look into my diff78 [May 11] (atomics support for SunOS x86/x64) and
diff88_cvs [May 28] (atomics fixes for gcc/msvc x64 and msvc x86).

libatomic_ops ChangeLog entries for diff88_cvs are in:
http://article.gmane.org/gmane.comp.programming.garbage-collection.boehmgc/2746

libatomic_ops ChangeLog entries for diff78:

        * src/atomic_ops/sysdeps/sunc/x86.h: New file.
        * src/atomic_ops/sysdeps/sunc/x86_64.h: New file.
        * src/atomic_ops.h (AO_INLINE): Define specially for Sun C.
        * src/atomic_ops.h: Fix comment (for x86_64).
        * src/atomic_ops.h: Include specialized x86.h and x86_64.h
        files for Sun C (if not AO_USE_PTHREAD_DEFS).

> I would like to make available another snapshot that includes the C++0x support, and is based on this snapshot.  But I appear to have broken that at the moment.

Are You going to have a standalone C++Ox distributive pack or to integrate C++0x support in bdwgc package? If the latter, are You going to "cvs" it?

> Other than that, I expect to not have time to work on this during the next week.  (If I introduced a really nasty bug in the repository, I'll try to deal with it, of course.)

Ok. I'll continue to resubmit my old patches (against the current CVS).
If You would introduce big changes to the CVS (before processing the sent patches), I could back review them.

> 
> Hans

Bye.


More information about the Gc mailing list