Defect: test 6 - allocate_as_barrier_proc (Failed) on Debian9/gcc-6.3/openmpi-3.1.0 #538

schloegl · 2018-05-19T20:31:13Z

When trying to install OpenCoarrays 2.0.0 on Debian 9 using gcc 6.3, and a openmpi 3.1.0, running make test failed with:

 5/49 Test  #5: allocate_as_barrier ..................   Passed    1.69 sec
      Start  6: allocate_as_barrier_proc
 6/49 Test  #6: allocate_as_barrier_proc .............***Failed  Required regular expression not found.Regex=[Test passed.
]  2.13 sec
      Start  7: get_array
 7/49 Test  #7: get_array ............................   Passed    0.95 sec
      Start  8: get_self
...
49/49 Test #49: test-installation-scripts.sh .........   Passed    0.46 sec

98% tests passed, 1 tests failed out of 49

Total Test time (real) =  44.68 sec

The following tests FAILED:
          6 - allocate_as_barrier_proc (Failed)
Errors while running CTest
Makefile:127: recipe for target 'test' failed
make: *** [test] Error 8

gcc/gfortran 6.3 from Debian 9 /stretch was was used for compiling OpenCorrarys as well as OpenMPI 3.1.0. gcc -v / gfortran -v report provide this report.

$ gfortran -v
Using built-in specs.
COLLECT_GCC=gfortran
COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-linux-gnu/6/lto-wrapper
Target: x86_64-linux-gnu
Configured with: ../src/configure -v --with-pkgversion='Debian 6.3.0-18+deb9u1' --with-bugurl=file:///usr/share/doc/gcc-6/README.Bugs --enable-languages=c,ada,c++,java,go,d,fortran,objc,obj-c++ --prefix=/usr --program-suffix=-6 --program-prefix=x86_64-linux-gnu- --enable-shared --enable-linker-build-id --libexecdir=/usr/lib --without-included-gettext --enable-threads=posix --libdir=/usr/lib --enable-nls --with-sysroot=/ --enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes --with-default-libstdcxx-abi=new --enable-gnu-unique-object --disable-vtable-verify --enable-libmpx --enable-plugin --enable-default-pie --with-system-zlib --disable-browser-plugin --enable-java-awt=gtk --enable-gtk-cairo --with-java-home=/usr/lib/jvm/java-1.5.0-gcj-6-amd64/jre --enable-java-home --with-jvm-root-dir=/usr/lib/jvm/java-1.5.0-gcj-6-amd64 --with-jvm-jar-dir=/usr/lib/jvm-exports/java-1.5.0-gcj-6-amd64 --with-arch-directory=amd64 --with-ecj-jar=/usr/share/java/eclipse-ecj.jar --with-target-system-zlib --enable-objc-gc=auto --enable-multiarch --with-arch-32=i686 --with-abi=m64 --with-multilib-list=m32,m64,mx32 --enable-multilib --with-tune=generic --enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu
Thread model: posix
gcc version 6.3.0 20170516 (Debian 6.3.0-18+deb9u1)

openmpi 3.1.0 was compiled from source with the following configuration flags:

       ./configure --prefix=/mnt/nfs/clustersw/$(lsb_release -is)/$(lsb_release -cs)/openmpi/${VER} \
                --enable-orterun-prefix-by-default  \
                --enable-mpi-cxx \
                --with-pmi \
                --with-pmix \
                --with-sge \
                --with-slurm \
                --with-pmi-libdir=/usr/lib/x86_64-linux-gnu  \

OpenCoarrays 2.0.0 was build with

        mkdir opencoarrays-build
        cd opencoarrays-build
        export FC=/mnt/nfs/clustersw/Debian/stretch/openmpi/3.1.0/bin/mpifort
        export CC=/mnt/nfs/clustersw/Debian/stretch/openmpi/3.1.0/bin/mpicc
        cmake ${BUILDDIR}/OpenCoarrays-${VER} \
          -DCMAKE_INSTALL_PREFIX=/mnt/nfs/clustersw/$(lsb_release -is)/$(lsb_release -cs)/${PKG}/${VER}
        make
        make test

This is were the failed tests (as described above) were observed.

Defect/Bug Report

OpenCoarrays Version: 2.0.0
Fortran Compiler: Debian, 6.3.0
C compiler used for building lib: 6.3.0
Installation method: cmake (see above)
Output of uname -a: Linux bea81 4.9.0-6-amd64 #1 SMP Debian 4.9.82-1+deb9u3 (2018-03-02) x86_64 GNU/Linux
MPI library being used: openmpi 3.1.0
Machine architecture and number of physical cores:x86_64, 24
Version of CMake: 3.7.2, preinstalled

Observed Behavior

The following tests FAILED:
          6 - allocate_as_barrier_proc (Failed)

98% tests passed, 1 tests failed out of 49

Total Test time (real) =  44.68 sec

The following tests FAILED:
          6 - allocate_as_barrier_proc (Failed)

Expected Behavior

   100% tests passed

Steps to Reproduce

installing openmpi 3.1.0 on Debian9, setting environment variables CC and FC to mpicc and mpifort, and compile OpenCoarrays 2.0.0 accordingly . Running make test will produce the failed test.

The text was updated successfully, but these errors were encountered:

schloegl · 2018-05-19T20:49:29Z

The above report is based on release 2.0.0 of OpenCoarrays.

The problem seems to be fixed in the repository (tested on commit 4d47cf2 ) . I can not say which commit fixed it. It would be good if an official release with this bug fixed would become available soon.

rouson · 2018-05-23T06:24:15Z

@schloegl Thanks for confirming that the fix is in the repository. That commit has been merged into the master branch and will therefore appear in the next release. Although we don't have an exact release schedule (or really even an inexact one!), we have several good reasons to release soon so the wait shouldn't be too long. There is a weekly OpenCoarrays videoconference that happens on Fridays so possibly we can get the next release out at the end of this week during or after the videoconference.

zbeekman · 2018-05-24T14:03:30Z

This may be a duplicate of #415. I still see non-deterministic test failures of allocate_as_barrier_proc every once in a while. I have no idea what the source is, so it's ~~likely~~ possible that you just hit this bug once and then it went away. ~~AFAIK, the have not been any/many commits to the library itself since 2.0.0.~~

EDIT: I lied, there are library changes so it IS possible that this bug (and #415) are now resolved. (Library code changes since 2.0.0: 2.0.0...master#diff-775a104746a19927fb80aad9b039981f)

schloegl · 2018-05-28T21:19:00Z

Seeing there is a new release 2.1.0, I've been trying to compile and test OpenCorrays again. Because, i've difficulties reproducing the earlier results, I run the tests also again on OpenCoarrays 2.0.0 (commit 4d47cf2). Moreover, I extended the tests gcc/gfortran 7.3.0 and 8.1.0, and openmpi-3.1.0 was recompiled with the corresponding compiler as well. The log files are attached

log-200.txt
log-210.txt

The log files how these results:

OpenCoarrays: 2.0.0 (commit `4d47cf2`)

gcc/gfortran 6.3.0:

6 - allocate_as_barrier_proc (Failed)

gcc/gfortran 7.3.0:

100% tests passed, 0 tests failed out of 67

gcc/gfortran 8.1.0:

compilation failed

OpenCoarrays: 2.1.0 (commit `8737658`)

gcc/gfortran 6.3.0

      6 - allocate_as_barrier_proc (Failed)
     51 - ISO_Fortran_binding_tests (Failed)

gcc/gfortran 7.3.0

     66 - ISO_Fortran_binding_tests (Failed)

gcc/gfortran 8.1.0

compilation failed

summary

the ISO_Fortran_binding_tests fails with OpenCoarrays 2.1.0.
the test allocate_as_barrier_proc fails on gcc/gfortran 6.3.0 but not on 7.3.0
It seems only gcc 7.3.0 with opencoarrays 2.0.0 has no failed tests.

zbeekman · 2018-07-01T01:00:57Z

I don't understand why ISO_Fortran_binding is failing for you if you're using GCC as the C compiler with 2.1.0. The other issues I believe I understand. (allocate_as_barrier_proc is fixed in GFortran >= 7.x and the issue with 8.1 appears to be due to a bug in the GFortran 8.1 release which requires this patch to fix.

allocate_as_barrier_proc will likely stay broken on the 6.x branch of GFortran, but I will turn off the test there since it's fixed on master. After PR #557 is merged can you try again (and this time run CTest with the --output-on-failure flag).

schloegl · 2018-07-03T23:48:08Z

I compiled from commit 8ef32a8 with gcc/7.3m openmpi/3.1.0, and the test failure "allocate_as_barrier_proc" did not appear. The test "#67: ISO_Fortran_binding_tests " still fails, attached is the error log produced with
CTEST_OUTPUT_ON_FAILURE=1 ctest

ctest.txt

zbeekman · 2018-07-04T00:07:08Z

We’re planning to make ISO_Fortran_binding it’s own project as it’s not required and only tangentially related to OpenCoarrays. The latest (still open) PR disables ISO_Fortran_binding by default. I haven’t been able to reproduce allocate_as_barrier_proc with 7.2 or 8.1 (patched). Is gcc/7.3m the 7.3 development branch? Can you compile the Debug configuration (`cmake /path/to/src -DCMAKE_BUILD_TYPE=Debug`) and run ctest with `--output-on-failure` copying the output of the failing tests? Thanks!

…

On Tue, Jul 3, 2018 at 7:48 PM Alois Schlögl ***@***.***> wrote: I compiled from commit 8ef32a8 <8ef32a8> with gcc/7.3m openmpi/3.1.0, and the test failure "allocate_as_barrier_proc" did not appear. The test "#67 <#67>: ISO_Fortran_binding_tests " still fails, attached is the error log produced with CTEST_OUTPUT_ON_FAILURE=1 ctest ctest.txt <https://github.com/sourceryinstitute/OpenCoarrays/files/2161383/ctest.txt> — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub <#538 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAREPGA4-fxbchdGRo_6U70iqXTXSnJxks5uDAK4gaJpZM4UF2E8> .

schloegl · 2018-07-04T16:01:52Z

I've been using the official release of gcc 7.3.0 (Jan 25, 2018, https://gcc.gnu.org/gcc-7/ ) build from source.

When using gcc 6.3, I'm using the version from debian/stretch https://tracker.debian.org/pkg/gcc-6, currently this is "gcc-Version 6.3.0 20170516 (Debian 6.3.0-18+deb9u1) ".

allocate_as_barrier_proc seems to fail only with gcc 6.3 (from debian/stable) but seems fine when compiling with gcc 7.3.

attached are the log files from cmake and ctest for gcc-7 as requested.

cmake.log
ctest.log

Since "allocate_as_barrier_proc will likely stay broken on the 6.x", perhaps this issue should be closed as the title refers to gcc-6.3, and perhaps a separate issued should be opened for the test "66 - ISO_Fortran_binding_tests (Failed)".

zbeekman · 2018-07-05T15:55:12Z

allocate_as_barrier_proc seems to fail only with gcc 6.3 (from debian/stable) but seems fine when compiling with gcc 7.3.

Great! As I said, I don't think we'll get around to fixing this on GCC 6, and as soon as we get some important GCC/GFortran 8 patches included in important GCC/GFortran *nix packages, we'll probably deprecate support for GCC 6.x and earlier.

Since "allocate_as_barrier_proc will likely stay broken on the 6.x", perhaps this issue should be closed as the title refers to gcc-6.3, and perhaps a separate issued should be opened for the test "66 - ISO_Fortran_binding_tests (Failed)".

We are aware of the problem with ISO_Fortran_binding and have a fix. Unless you object I'm going to close this issue because ISO_Fortran_binding is tracked elsewhere (#554 and others)

zbeekman · 2018-07-05T15:55:48Z

If you disagree, with closing this, or encounter a new issue feel free to reopen or post a new issue.

schloegl closed this as completed May 19, 2018

schloegl reopened this May 28, 2018

zbeekman changed the title ~~test 6 - allocate_as_barrier_proc (Failed) on Debian9/gcc-6.3/openmpi-3.1.0~~ Defect: test 6 - allocate_as_barrier_proc (Failed) on Debian9/gcc-6.3/openmpi-3.1.0 Jul 1, 2018

zbeekman self-assigned this Jul 1, 2018

zbeekman added bug duplicate labels Jul 1, 2018

zbeekman closed this as completed Jul 5, 2018

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Defect: test 6 - allocate_as_barrier_proc (Failed) on Debian9/gcc-6.3/openmpi-3.1.0 #538

Defect: test 6 - allocate_as_barrier_proc (Failed) on Debian9/gcc-6.3/openmpi-3.1.0 #538

schloegl commented May 19, 2018 •

edited by zbeekman

Loading

schloegl commented May 19, 2018

rouson commented May 23, 2018

zbeekman commented May 24, 2018 •

edited

Loading

schloegl commented May 28, 2018 •

edited by zbeekman

Loading

zbeekman commented Jul 1, 2018

schloegl commented Jul 3, 2018

zbeekman commented Jul 4, 2018 via email

schloegl commented Jul 4, 2018

zbeekman commented Jul 5, 2018

zbeekman commented Jul 5, 2018

Defect: test 6 - allocate_as_barrier_proc (Failed) on Debian9/gcc-6.3/openmpi-3.1.0 #538

Defect: test 6 - allocate_as_barrier_proc (Failed) on Debian9/gcc-6.3/openmpi-3.1.0 #538

Comments

schloegl commented May 19, 2018 • edited by zbeekman Loading

Defect/Bug Report

Observed Behavior

Expected Behavior

Steps to Reproduce

schloegl commented May 19, 2018

rouson commented May 23, 2018

zbeekman commented May 24, 2018 • edited Loading

schloegl commented May 28, 2018 • edited by zbeekman Loading

OpenCoarrays: 2.0.0 (commit 4d47cf2)

gcc/gfortran 6.3.0:

gcc/gfortran 7.3.0:

gcc/gfortran 8.1.0:

OpenCoarrays: 2.1.0 (commit 8737658)

gcc/gfortran 6.3.0

gcc/gfortran 7.3.0

gcc/gfortran 8.1.0

summary

zbeekman commented Jul 1, 2018

schloegl commented Jul 3, 2018

zbeekman commented Jul 4, 2018 via email

schloegl commented Jul 4, 2018

zbeekman commented Jul 5, 2018

zbeekman commented Jul 5, 2018

schloegl commented May 19, 2018 •

edited by zbeekman

Loading

zbeekman commented May 24, 2018 •

edited

Loading

schloegl commented May 28, 2018 •

edited by zbeekman

Loading

OpenCoarrays: 2.0.0 (commit `4d47cf2`)

OpenCoarrays: 2.1.0 (commit `8737658`)