Open MPI logo

Open MPI Development Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Development mailing list

Subject: [OMPI devel] 1.7.0rc5 - make check failure on OpenBSD-5.1/{i386, amd64}
From: Paul Hargrove (phhargrove_at_[hidden])
Date: 2012-10-30 23:25:18


My OpenBSD-5.1 testers for both i386 and amd64 are failing the same tests
as I reported a few minutes ago with respect to FreeBSD-6.3.

Unlike FreeBSD-6.3, this is "modern" system with OpenBSD 5.1 having been
released Feb 2012.

On both platforms I have builds w/ gcc-4.2.1 and with llvm-gcc-4.2.1.
All 4 combinations of arch/compiler SEGV on the same 3 datatype tests.

UNLIKE the FreeBSD-6.3 failures, the backtraces (though still numeric) have
only a depth of 10 (at least in the cases I sampled at random in my limited
time).

FWIW: NetBSD-5.1 looks fine on both i386 and amd64 h/w.

-Paul

On Tue, Oct 30, 2012 at 8:13 PM, Paul Hargrove <phhargrove_at_[hidden]> wrote:

> On my FreeBSD-6.3/amd64 platform I see "make check" failing 3 tests under
> test/datatype (see below). Of course "make" stops after that, making it
> possible that additional tests might fail later.
>
> However, my records do show that the v1.5 branch was just fine on this
> machine, as was the trunk on or about Feb 21, 2012. So, unless the failing
> tests are new THIS IS A REGRESSION.
>
> Newer FreeBSD-7.2/amd64, FreeBSD-8.2/amd64 and FreeBSD-9.0/amd64 are all
> OK.
>
>
> The failures:
>
> $ grep -B1 ^FAIL $HOME/OMPI/openmpi-1.7rc5-freebsd6-amd64/LOG/check.log
> Segmentation fault (core dumped)
> FAIL: position
> --
> Segmentation fault (core dumped)
> FAIL: ddt_test
> Segmentation fault (core dumped)
> FAIL: ddt_raw
>
> Using gdb to get a backtrace from the generated core files shows a stack
> over 1300 frames deep, which I suspect is either evidence of a corrupted
> stack (buffer overflow?) or of infinite recursion. There are no symbols
> with the backtrace, just numeric addresses.
>
> This is a rather old system
> $ gcc --version | head -1
> gcc (GCC) 3.4.6 [FreeBSD] 20060305
>
>
> -Paul
>
>
> On Tue, Oct 30, 2012 at 7:01 PM, Ralph Castain <rhc_at_[hidden]> wrote:
>
>> Hi folks
>>
>> We have posted the next release candidate (rc5) for the 1.7.0 release in
>> the usual place:
>>
>> http://www.open-mpi.org/software/ompi/v1.7/
>>
>> Please put it thru the wringer to help us validate it prior to release
>> later this month. We think this looks pretty complete, pending someone
>> finding a problem.
>>
>> Thanks
>> Ralph
>>
>> _______________________________________________
>> devel mailing list
>> devel_at_[hidden]
>> http://www.open-mpi.org/mailman/listinfo.cgi/devel
>>
>
>
>
> --
> Paul H. Hargrove PHHargrove_at_[hidden]
> Future Technologies Group
> Computer and Data Sciences Department Tel: +1-510-495-2352
> Lawrence Berkeley National Laboratory Fax: +1-510-486-6900
>
>

-- 
Paul H. Hargrove                          PHHargrove_at_[hidden]
Future Technologies Group
Computer and Data Sciences Department     Tel: +1-510-495-2352
Lawrence Berkeley National Laboratory     Fax: +1-510-486-6900