gcc 4.2.0, r120684 backport and Process_1/Process_5

Jack Howarth howarth@bromo.msbb.uc.edu
Thu May 17 16:04:00 GMT 2007


David,
   At the moment I am assuming that r120684 requires
some additional patches. Specifically, I noticed
that...

r120801 | andreast | 2007-01-15 16:24:42 -0500 (Mon, 15 Jan 2007) | 6 lines

2007-01-15  Andreas Tobler  <a.tobler@schweiz.org>

        * os_dep.c (defined(MPROTECT_VDB) && defined(DARWIN)): Adjust mail
        reference.
        (catch_exception_raise): Fix typo in the I386 exc_state.

r120853 | andreast | 2007-01-17 07:01:45 -0500 (Wed, 17 Jan 2007) | 4 lines

2007-01-17  Andreas Tobler  <a.tobler@schweiz.org>

        * include/gc_config.h.in: Regenerate.

r120874 | mrs | 2007-01-17 15:12:51 -0500 (Wed, 17 Jan 2007) | 2 lines

        * os_dep.c: Fix i686-apple-darwin9 builds.

all seem to be cleanups associated with r120684. I am going to
try a build on powerpc-apple-darwin8 and i686-apple-darwin8
tonight with r120684, r120801, r120853 and r120874 backported
to gcc 4.2.0 and see if that eliminates the random Process_#
tests failures and reduces the libjava failures further on
Macintel.
            Jack

On Thu, May 17, 2007 at 08:43:27AM -0700, David Daney wrote:
> Jack Howarth wrote:
> >    Can anyone tell me what the expected testsuite results
> >are for libjava on i686-apple-darwin8 with current gcc trunk?
> 
> I always expect zero FAILs, but I have never run the testsuite on 
> i686-apple-darwin8 either.
> 
> >I backported the changes in r120684 to the release gcc 4.2.0...
> >
> >http://gcc.gnu.org/ml/gcc/2007-05/msg00437.html
> >
> >and it reduced the testsuite failures from 90 to 27. However I now
> >see a single new failure for Process_5. On powerpc-apple-darwin8
> >the same changes introduces a single failure for Process_1 instead.
> >Interestingly on powerpc-apple-darwin8,
> 
> The process tests rely on a couple of things.  The first is that signals 
> and threads are posix like.  The second is that 'sleep' and 'echo' are 
> in the PATH and have posix semantics.
> 
> You could run the tests by hand and report the output.  That would help 
> us better understand what is happening.
> 
> David Daney



More information about the Java mailing list