[Bug libgomp/46967] New: lots of testsuite failures with libgomp on hppa-hp-hpux11.31

[Bug libgomp/46967] New: lots of testsuite failures with libgomp on hppa-hp-hpux11.31

[bugzilla-gcc at thewrittenword dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46967

           Summary: lots of testsuite failures with libgomp on
                    hppa-hp-hpux11.31
           Product: gcc
           Version: 4.4.5
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libgomp
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: bugzilla-gcc@thewrittenword.com


Running /opt/build/gcc-4.4.5/libgomp/testsuite/libgomp.c/c.exp ...
FAIL: libgomp.c/copyin-1.c execution test
FAIL: libgomp.c/copyin-3.c execution test
FAIL: libgomp.c/nested-2.c execution test
FAIL: libgomp.c/pr24455.c execution test
FAIL: libgomp.c/pr29947-2.c execution test
FAIL: libgomp.c/pr42942.c execution test
Running /opt/build/gcc-4.4.5/libgomp/testsuite/libgomp.c++/c++.exp ...
Pid 23693 in trap loop, signal 11
[snip]
Pid 771 in trap loop, signal 11
FAIL: libgomp.c++/ctor-9.C  -O3 -fomit-frame-pointer -funroll-loops
execution test
FAIL: libgomp.c++/for-1.C  -O1  execution test
FAIL: libgomp.c++/for-1.C  -O2  execution test
FAIL: libgomp.c++/for-1.C  -O3 -fomit-frame-pointer  execution test
Pid 930 in trap loop, signal 11
FAIL: libgomp.c++/for-1.C  -O3 -fomit-frame-pointer -funroll-loops
execution test
[snip]
Pid 15005 in trap loop, signal 11
FAIL: libgomp.fortran/task1.f90  -O0  execution test
Pid 15641 in trap loop, signal 11
Test Run By pogma on Thu Dec  2 19:18:38 2010
Native configuration is hppa2.0-hp-hpux11.31
                === libgomp Summary ===

# of expected passes            2285
# of unexpected failures        173
# of unsupported tests          9


HP-UX 11.23 doesn't have any problems with the libgomp tests.

See also my email to gcc@ asking for help diagnosing this:
http://gcc.gnu.org/ml/gcc/2010-12/msg00116.html

[Bug libgomp/46967] lots of testsuite failures with libgomp on hppa-hp-hpux11.31

[bugzilla-gcc at thewrittenword dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46967

--- Comment #1 from The Written Word <bugzilla-gcc at thewrittenword dot com> 2011-01-27 20:06:40 UTC ---

John David Anglin pointed us to bug
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34544 so we backported the patch to
4.4 and tried it.

 With r163461 backported (compiler only built with --enable-languages=c this
time) we ran the libgomp testsuite 3 times, failed 11/248 tests, 0/248 tests
and 9/248 tests. So, still not fantastic, but the 0/248 was nice to see.

However, the big win for us is that perl-5.12.2 built with the compiler with
the backported patch fails no tests at all where it used to fail many.

[Bug libgomp/46967] lots of testsuite failures with libgomp on hppa-hp-hpux11.31

[dave at hiauly1 dot hia.nrc.ca]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46967

--- Comment #2 from dave at hiauly1 dot hia.nrc.ca 2011-01-27 20:24:02 UTC ---
>  With r163461 backported (compiler only built with --enable-languages=c this
> time) we ran the libgomp testsuite 3 times, failed 11/248 tests, 0/248 tests
> and 9/248 tests. So, still not fantastic, but the 0/248 was nice to see.

I believe that libgomp.sl is incorrectly linked:

# chatr libgomp.sl
libgomp.sl:
         shared library
     shared library dynamic path search:
         SHLIB_PATH     disabled  second
         embedded path  enabled   first  /opt/gnu/gcc/gcc-4.5/lib
     internal name:
         libgomp.sl.1
     shared library list:
         dynamic   /usr/lib/librt.2
         dynamic   /usr/lib/libc.2
         dynamic   /mnt/gnu/gcc/objdir/./gcc/libgcc_s.4
         dynamic   /opt/langtools/lib/libpthread.1

As can be seen, libgomp is linked against libc, and libc.2 comes
before libpthread.1 in the shared library list.  libc contains
non functional pthread functions, so it is likely pthread support
is not functional.  From the discussion, it appears there is a
difference in the behavior of the dynamic loader in 11.31 and
other hpux 11 releases.

HP recommends not linking shared libraries against libc.  libtool
adds -lc to the link command.  It doesn't come directly from the libgomp
configuration.  So, libtool needs to be fixed first.

Dave

[Bug libgomp/46967] lots of testsuite failures with libgomp on hppa-hp-hpux11.31

[htl10 at users dot sourceforge.net]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46967

Hin-Tak Leung <htl10 at users dot sourceforge.net> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |htl10 at users dot
                   |                            |sourceforge.net

--- Comment #3 from Hin-Tak Leung <htl10 at users dot sourceforge.net> 2011-05-01 19:01:17 UTC ---
Hmm, I was looking around for other bugs and wondering if I should fail one -
just finished testing 4.4.6 on alphaev68-dec-osf5.1a and 

(my own previous result, on the same machine):
http://gcc.gnu.org/ml/gcc-testresults/2010-12/msg01338.html

4.4.5:
                === libgomp Summary ===

# of expected passes            2463
# of unexpected failures        9
# of unsupported tests          2

and the new one in 4.4.6 (I'll post to gcc-testresults@gcc soon):

         === libgomp Summary ===

# of expected passes        1610
# of unexpected failures    466
# of unsupported tests        2

so something has gone very wrong. with 4.4.6.

Since HP owns Tru64 these days, I'd be interested in hppa-hp-hpux11.31's 4.4.6
's test result.

[Bug libgomp/46967] lots of testsuite failures with libgomp on hppa-hp-hpux11.31

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46967

John David Anglin <danglin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |WAITING
   Last reconfirmed|                            |2011.05.01 21:28:15
     Ever Confirmed|0                           |1

--- Comment #4 from John David Anglin <danglin at gcc dot gnu.org> 2011-05-01 21:28:15 UTC ---
This should be fixed.  Albert, can you confirm that the pthread active
changes fixed this problem?

Regarding comment #3, look at the libgomp test log file to see why the
tests are failing on alphaev68-dec-osf5.1a.  I'm certain the problem
is different from the hppa2.0w-hp-hpux11.31 issue, so a new PR
should be opened.

[Bug libgomp/46967] lots of testsuite failures with libgomp on hppa-hp-hpux11.31

[htl10 at users dot sourceforge.net]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46967

--- Comment #5 from Hin-Tak Leung <htl10 at users dot sourceforge.net> 2011-05-01 22:02:03 UTC ---
(In reply to comment #4)
> Regarding comment #3, look at the libgomp test log file to see why the
> tests are failing on alphaev68-dec-osf5.1a.  I'm certain the problem
> is different from the hppa2.0w-hp-hpux11.31 issue, so a new PR
> should be opened.

Will do. am checking bug 40894 for 4.4.6 at the moment. When I finish updating
the status of that I'll send the test result to test-results, and file  a new
bug.

[Bug libgomp/46967] lots of testsuite failures with libgomp on hppa-hp-hpux11.31

[htl10 at users dot sourceforge.net]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46967

--- Comment #6 from Hin-Tak Leung <htl10 at users dot sourceforge.net> 2011-05-01 22:24:08 UTC ---
Filed Bug 48841 for the alphaev68 libgomp failure and attached my test summary,
in case somebody wants to compare to hppa 4.4.6.

[Bug libgomp/46967] lots of testsuite failures with libgomp on hppa-hp-hpux11.31

[bugzilla-gcc at thewrittenword dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46967

--- Comment #7 from The Written Word <bugzilla-gcc at thewrittenword dot com> 2011-08-20 03:34:48 UTC ---
(In reply to comment #4)
> This should be fixed.  Albert, can you confirm that the pthread active
> changes fixed this problem?
> 
> Regarding comment #3, look at the libgomp test log file to see why the
> tests are failing on alphaev68-dec-osf5.1a.  I'm certain the problem
> is different from the hppa2.0w-hp-hpux11.31 issue, so a new PR
> should be opened.

It should be fixed in 4.4.6? What revision fixed the problem?

[Bug libgomp/46967] lots of testsuite failures with libgomp on hppa-hp-hpux11.31

[dave.anglin at bell dot net]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46967

--- Comment #8 from dave.anglin at bell dot net 2011-08-20 15:45:46 UTC ---
The pthread change was applied in r170290.  If it doesn't fix the bug,  
then further
debugging is needed.  I can say the tests don't fail with 11.11.  You  
also indicated
the tests don't fail with 11.23.  So, these fails either have to do  
with differences in
linking, or processor architecture.

The linking issue mentioned in comment #2 is a libtool issue.  It  
incorrectly links in libc
in shared library links.  It is possible to hack libtool to not link  
in libtool on hppa*-*-hpux*.
Some other targets do this, so there is sample code.  If the non  
functional pthread
stubs are used, races will result causing test failures.  The for-*.C  
tests are especially
sensitive to this as they create many threads to parallelize loops.

The libtool project is essentially dormant, and none of the  
maintainers have responded
to my patches or bug reports in recent months.

I believe there are subtle differences in the behavior of the dynamic  
linker in different
hpux versions.

There is also the issue that 11.31 usually only runs on PA8800/PA8900  
servers.  These
CPUs have much larger instruction and data caches (32 MB), and the  
caches do not
support non equivalent aliasing.  From experience, I know that it is  
much more difficult
to do a good pthread implementation on these CPUs.   More cache  
flushing is needed
than on earlier CPUs, and this changes program timing quite a bit.   
It's quite tricky to
avoid the aliasing issues and cache corruption.

--
John David Anglin    dave.anglin@bell.net

[Bug libgomp/46967] lots of testsuite failures with libgomp on hppa-hp-hpux11.31

[danglin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=46967

John David Anglin <danglin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |RESOLVED
         Resolution|---                         |WONTFIX

--- Comment #9 from John David Anglin <danglin at gcc dot gnu.org> ---
Closing as 4.4 is no longer supported.