MOOSE Test Failures (575 Tests FAILED)!

[MT]

Hi All,

I was installing MOOSE for the first time and apparently ran into some issues during the install.  I have a MacBook Pro running Snow Leopard (I am not allowed to update the software to Yosemite) and I was attempting to manually install MOOSE on OS X manually using the Basic Manual Installation Guide (http://mooseframework.org/wiki/BasicManualInstallation/OSX/).  I didn't encounter any errors and installed Peacock.  In Section 6 located in the previous link it states, " recommend that you browse your $PACKAGE_DIR directory and make sure that you see each of the library directories in there (OpenMPI, Hypre, PETSc)."  I checked $PACKAGE_DIR and only found OpenMPI and PETSc directories... I followed the directions several times to the "T" and still, only the OpenMPI and PETSc directories are located in the $PACKAGE_DIR.  I continued with the install following Sections 2, 3, and 4 in the following hyperlink (http://mooseframework.com/getting-started/) without any errors until I began testing the software.  I have attached the output Terminal feedback... Out of 957 tests; 382 passed, 1 skipped, 0 pending, 575 FAILED.

Could someone lend a hand and shed some light on why I am getting so many failures?

I greatly appreciate your help; thank you in advance!

MT

[Andrew....@csiro.au]

Hi MT,

I cannot help you directly, but thought I’d mention that apparently all of the failures you obtained from the tests I wrote are in fact from input files for which MOOSE is *supposed to crash*.  They’re mostly testing things like invalid input files (where the user has screwed up and written an erroneous input file), and the test harness is supposed to realise this, and say “test passed” because MOOSE crashed in the correct place.  That suggests to me that your test harness is not working correctly, but possibly moose itself is fine.

a

--
You received this message because you are subscribed to the Google Groups "moose-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to moose-users...@googlegroups.com.
Visit this group at http://groups.google.com/group/moose-users.
To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/6a1bb058-e1bf-437c-bd34-2ea621f80838%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[MT]

Thanks for your thoughts... I initially thought the same thing at first but the MOOSE site states, "If everything is good then all of the tests will pass. Note: Some will be skipped depending on your environment."

I feel like the build failed but I followed all the directions... any thoughts on how to fix this?

MT

[Jason Miller]

Looks like the context of those instructions could use an update. We no longer build Hypre. We let PETSc do that for us. Hence, you will not see a `hypre` directory. But other than that, you probably built everything just fine. When a large amount of tests fail, that usually means somethings wrong with the environment. Be it a wrong environment variable, or some other library getting in our way (MacPorts/Fink/HomeBrew comes to mind).

Please post the results of an `env` command, so that we might have a better understanding of whats whats.

env > my_environment.txt

email me that my_environment.txt file.

Thanks!

Jason Miller

To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/cba1d7c6-30c0-4965-8cbb-dc9509168f95%40googlegroups.com.

[Jean Francois Leon]

Hi All
I went through the same experience as MT above [ in advance of the forthcoming workshop next week] and ran into the same issues.
Followed step by step instructions from moose website to install and setup moose and test it. [on opensuse 13.2]
No error or warning popped up. everything went smoothly apparently
But when I ran the test here is what I got:

Ran 922 tests in 0.5 seconds
0 passed, 7 skipped, 0 pending, 922 FAILED

most of the failure were
.......... Max Fails Exceeded
and a few of them toward the end just
FAILED (CRASH) ( in red)

Do i need to do anything to fix that?
Thanks
JF

[MT]

I have attached the "env" output; thanks for looking into this!

MT

[Cody Permann]

MT,

Jason has loaded Snow Leopard on a system and will see if he can replicate. We may or may not be able to fix this issue. Neither we nor Apple supports that operating system at this point. It has several problems including broken compilers (not an exaggeration). We'll let you know if we can come up with a fix but we can't expend too many resources on an unsupported system.

Cody

To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/5661ceb8-6c98-4c47-9434-5990674fb338%40googlegroups.com.

[Cody Permann]

Jean,

It's possible that one of your steps failed and you just didn't see the message. Some of the steps don't produce really obvious messages so they are easy to miss. I suggest you type "make" once more and see if there's an error. If there's nothing to do (success), the make system will tell you that.

Let us know what you see.

Cody

--

You received this message because you are subscribed to the Google Groups "moose-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to moose-users...@googlegroups.com.
Visit this group at http://groups.google.com/group/moose-users.

To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/9d15323b-dcd2-4c45-b940-6588417ecdfb%40googlegroups.com.

[Jean Francois Leon]

Hi Cody
Thanks for your prompt answer
Here is what I see now [ second time different from first one but indeed an erro]r:
jf@linux:~/projects/moose/test> make -j8
Linking Library /home/jf/projects/moose/framework/libmoose-opt.la...
/usr/bin/grep: /home/jf/projects/moose/scripts/../libmesh/installed/lib64/libmesh_opt.la: No such file or directory
/usr/bin/sed: can't read /home/jf/projects/moose/scripts/../libmesh/installed/lib64/libmesh_opt.la: No such file or directory
libtool: link: `/home/jf/projects/moose/scripts/../libmesh/installed/lib64/libmesh_opt.la' is not a valid libtool archive
/home/jf/projects/moose/framework/moose.mk:75: recipe for target '/home/jf/projects/moose/framework/libmoose-opt.la' failed
make: *** [/home/jf/projects/moose/framework/libmoose-opt.la] Error 1

I thought libmesh compile went smoothly though..
Amy suggestion on what I can do to fix it?
Thanks
JF
Thanks
JF

[Cody Permann]

Yeah the libMesh script does a lot... I've thought about changing it around a bit to notify users of the failure with a better message. Try re-running the libMesh script and lets take a look at the output. Look for the word "ERROR" or "FAILURE" in there.

Cody

To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/b6bd5def-5435-45ec-aaa3-f7c9e2effcbe%40googlegroups.com.

[Jean Francois Leon]

I reran the libmesh script and there is no ERROR or FAILURE reported (case sensitive search as there are many filenames with error in them). Just  a few warnings.
I have captured the script terminal output in a text file, it is ~ 500K.I can sent it if you want.
JF

[Cody Permann]

I guess you'd better send the output. If that script succeeds, it'll build the libmesh library and put it in moose/libmesh/lib, but that didn't happen as evidence of your previous error report (missing library libmesh-opt.la....).

Cody

To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/efa90f61-6397-407e-9a85-87963a0b3392%40googlegroups.com.

[Jean Francois Leon]

done ( privately ) not too clutter the list with a large file
jf

[Miller, Jason M]

I would like a copy of that file. Just so I can follow along and throw my two cents at it.

I am still installing/prep`ing Snow Leopard. It has been quite an adventure gathering everything I need to start building on it. Not to mention finding a machine capable of running Snow Leopard.

To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/b0c9ac3f-c02f-43b4-87f2-2ff7e6a67348%40googlegroups.com.

[Miller, Jason M]

I ran through the Linux instructions on a brand new installation of OpenSUSE

http://mooseframework.org/wiki/BasicManualInstallation/Linux/

Before I started with those instructions, I made sure to `sudo zypper update` until there were no more updates.

Final results:

Ran 929 tests in 1415.3 seconds
929 passed, 19 skipped, 0 pending, 0 failed
moose@inl420736:~/projects/moose/test> lsb_release
LSB Version:    n/a
moose@inl420736:~/projects/moose/test> lsb_release -a
LSB Version:    n/a
Distributor ID: openSUSE project
Description:    openSUSE 13.2 (Harlequin) (x86_64)
Release:        13.2
Codename:       Harlequin

Something I want to point out: Running the tests in parallel was causing the TestHarness to quit at random. I am not sure why. Example:

...
...
functions/parsed.vector................................................................................... OK
functions/parsed.scalar................................................................................... OK
functions/piecewise_constant.test......................................................................... OK
functions/parsed.transient................................................................................ OK
executioners/transient_sync_time.testsynctimes............................................................ OK

[1]+  Stopped                 ./run_tests -j 4

However, running the tests serially ( ./run_tests -j 1) works. Why you see that large time (Ran 929 tests in 1415.3 seconds)

MOOSE hash (version) I was working with:

moose@inl420736:~/projects/moose/test/tests/kernels/simple_diffusion> ../../../moose_test-opt -i simple_diffusion.i
Framework Information:
MOOSE version:           git commit 8631d60 on 2015-06-19
PETSc Version:           3.5.2
Current Time:            Fri Jun 19 16:12:03 2015
Executable Timestamp:    Fri Jun 19 15:20:39 2015

Want to point out, at this point Peacock also worked. If it didn't work, you would see lots of problems right away:

moose@inl420736:~/projects/moose/test/tests/kernels/simple_diffusion> ../../../../gui/peacock -i simple_diffusion.i
Using Executable: /home/moose/projects/moose/test/moose_test-opt

     (  you'll have to take my word for it. Peacock popped open on my screen ;)  )

Here is moose_test-opt's links, as this might shed some light on your issue (please perform an `ldd moose_test-opt` on your machine):


moose@inl420736:~/projects/moose/test> ldd moose_test-opt
        linux-vdso.so.1 (0x00007ffe2135f000)
        libmoose_test-opt.so.0 => /home/moose/projects/moose/test/lib/libmoose_test-opt.so.0 (0x00007fad4611d000)
        libmoose-opt.so.0 => /home/moose/projects/moose/framework/libmoose-opt.so.0 (0x00007fad45260000)
        libpcre-opt.so.0 => /home/moose/projects/moose/framework/contrib/pcre/libpcre-opt.so.0 (0x00007fad45033000)
        libstdc++.so.6 => /usr/lib64/libstdc++.so.6 (0x00007fad44d0d000)
        libgcc_s.so.1 => /lib64/libgcc_s.so.1 (0x00007fad44af6000)
        libpthread.so.0 => /lib64/libpthread.so.0 (0x00007fad448d9000)
        libmesh_opt.so.0 => /home/moose/projects/moose/scripts/../libmesh/installed/lib64/libmesh_opt.so.0 (0x00007fad43781000)
        libnetcdf.so.7 => /home/moose/projects/moose/scripts/../libmesh/installed/lib64/libnetcdf.so.7 (0x00007fad43499000)
        libpetsc.so.3.5 => /opt/moose/petsc/openmpi_petsc-3.5.2/gcc-opt/lib/libpetsc.so.3.5 (0x00007fad42501000)
        libparmetis.so => /opt/moose/petsc/openmpi_petsc-3.5.2/gcc-opt/lib/libparmetis.so (0x00007fad422c3000)
        libmetis.so => /opt/moose/petsc/openmpi_petsc-3.5.2/gcc-opt/lib/libmetis.so (0x00007fad42058000)
        libmpi_usempi.so.1 => /opt/moose/openmpi/openmpi-1.8.1/gcc-opt/lib64/libmpi_usempi.so.1 (0x00007fad41d5e000)
        libmpi_mpifh.so.2 => /opt/moose/openmpi/openmpi-1.8.1/gcc-opt/lib64/libmpi_mpifh.so.2 (0x00007fad41b04000)
        libgfortran.so.3 => /usr/lib64/libgfortran.so.3 (0x00007fad417ea000)
        libquadmath.so.0 => /usr/lib64/libquadmath.so.0 (0x00007fad415ae000)
        libgomp.so.1 => /usr/lib64/libgomp.so.1 (0x00007fad4139e000)
        libmpi_cxx.so.1 => /opt/moose/openmpi/openmpi-1.8.1/gcc-opt/lib64/libmpi_cxx.so.1 (0x00007fad41183000)
        libmpi.so.1 => /opt/moose/openmpi/openmpi-1.8.1/gcc-opt/lib64/libmpi.so.1 (0x00007fad40ea9000)
        libopen-rte.so.7 => /opt/moose/openmpi/openmpi-1.8.1/gcc-opt/lib64/libopen-rte.so.7 (0x00007fad40c2c000)
        libopen-pal.so.6 => /opt/moose/openmpi/openmpi-1.8.1/gcc-opt/lib64/libopen-pal.so.6 (0x00007fad4094d000)
        libdl.so.2 => /lib64/libdl.so.2 (0x00007fad40749000)
        librt.so.1 => /lib64/librt.so.1 (0x00007fad40540000)
        libnsl.so.1 => /lib64/libnsl.so.1 (0x00007fad40328000)
        libutil.so.1 => /lib64/libutil.so.1 (0x00007fad40125000)
        libm.so.6 => /lib64/libm.so.6 (0x00007fad3fe23000)
        libc.so.6 => /lib64/libc.so.6 (0x00007fad3fa7c000)
        /lib64/ld-linux-x86-64.so.2 (0x00007fad4665e000)

Items in my /opt/moose directory:


moose@inl420736:~/projects/moose/test> ls -latr /opt/moose
total 0
drwxr-xr-x 1 root root  26 Jun 19 12:47 openmpi
drwxr-xr-x 1 root root  10 Jun 19 12:47 ..
drwxr-xr-x 1 root root  38 Jun 19 13:32 petsc
drwxr-xr-x 1 root root  42 Jun 19 13:35 .
drwxr-xr-x 1 root root 170 Jun 19 13:54 miniconda

The things that were wrong with the Linux instructions (and I left them alone for sake of keeping this email legitimate):

Installing Miniconda is incorrect. $PACKAGES_DIR/stack_src does not exist. That bit of instruction was pulled from another set of instructions I use for building the RPM, DEB and PKG packages.

I needed to `zypper install` several more packages before I could begin: git, m4

libMesh failed to build due to missing m4 (hence I had to go back, `zypper install m4`, rebuild libMesh). I have been told that is fairly random though.

The length of this email is not meant to be offensive! Just trying to offer up a lot of "here's how it should look" stuff. I am real curious about the results of your `ldd moose_test-opt` though! Please submit that here!

Thanks, and I hope some of this helps!

Jason

[Jean Francois Leon]

Jason
Thanks
I sent you the file privately
I am going to go through your email and check if I can id any discrepancies with my configuration.
I did not do the manual install but followed the 4 steps process exactely as described:
http://mooseframework.org/getting-started/
I must point that this is a laptop installation I am making for the workshop but that I did the same install a few months ago on my workstation [ same OS] and it worked "out of the box"
Thanks again
JF

[Jean Francois Leon]

Hi All
I followed Jason's route [ manual install on opensuse 13.2 and (almost) reproduced his result [ so I have a working configuration now].

Install went smoothly except for some peacock related stuff:
vtk installed with a broken dependency reported and yaml nowhere to be found. I guess I can live without peacock anyway at least for now.

test
jj  1:

 Ran 929 tests in 1126.2 seconds

929 passed, 19 skipped, 0 pending, 0 failed

and with j 4:

Ran 929 tests in 463.4 seconds

929 passed, 19 skipped, 0 pending, 0 failed

JF

[Cody Permann]

Well that's better news. You should be able to install both the VTK pieces and Yaml pieces you need through SuSE's package managers. Jason may have other ideas for you.

Cody

--
You received this message because you are subscribed to the Google Groups "moose-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to moose-users...@googlegroups.com.
Visit this group at http://groups.google.com/group/moose-users.

To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/a5338091-48fd-4ba8-a384-1f4b16c1261d%40googlegroups.com.

[Miller, Jason M]

Good to hear!

Because SuSE allows a choice of desktop flavor, which one did you choose? The default is KDE, so I have been curbing my instructions that direction. KDE requires Qt. As does our implementation of VTK for use with Peacock. So if you choose Gnome for example, might be why somethings are still missing. Nevertheless, you have the hard part done. As Cody mentions, installing VTK and Yaml are simple affairs through your package manager.

I believe:

  sudo -E zypper install python-PyYAML vtk

will do it. You'll probably run into a few more after you get those installed... matplotlib, numpy comes to mind.

I do have it listed, that you 'can' install these things with Miniconda and just be done with it. But it is best to try and install via your package manager first.

See you at the workshop! You guys can all yell at me there for my lousy instructions =D

Jason

[Jean Francois Leon]

Yes I use gnome indeed

Thanks for the pointer

Jf

You received this message because you are subscribed to a topic in the Google Groups "moose-users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/moose-users/H6r0MmgqtHg/unsubscribe.
To unsubscribe from this group and all its topics, send an email to moose-users...@googlegroups.com.

Visit this group at http://groups.google.com/group/moose-users.

To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/CALgZTAdFpEr26zLaJ%3Dr-5%2BHvFyu%3Dr4q6tAgsOiMZRCBJitt-Fg%40mail.gmail.com.

For more options, visit https://groups.google.com/d/optout.

--

Jean-François Leon
Galtenco

Tel [Main-Global]: +1 (415) 322-0376
Cell [France]: +33 (651) 168-014

[Miller, Jason M]

MT,

Concerning Snow Leopard,

I have followed the instructions located here:

http://mooseframework.org/wiki/BasicManualInstallation/OSX/

And have successfully tested the latest version of a MOOSE clone. I have one test failing (preconditioners/pbp.lots_of_variables due to a Time Out... this is a very old machine so in all fairness, thats normal)

Some things I ran into during build that were not part of the install instructions:

Snow Leopard's Xcode does not come with 'git'. I had to manually build that from source.

The latest version of CMake, is a bit difficult to source correctly. At the time of writing those instructions, it was a simple double click package installer.

With those items in mind, how did you solve those issues?

Things I was shocked to find out:

Building GCC 4.9.1 went flawlessly. Took a day. But it built. On the first try.

Edit:

I noticed that I did not install Miniconda, as I was not going to test for Peacock. However, after I decided I had better install Miniconda since its on the list of things to do!  I re-ran the update_and_rebuild_libmesh.sh, make cleanall; make... I am getting the exact same errors as you. So... cool! I have duplicated your issue. The good news is, I know this _can_work. Cause I did get only 1 failing test.

I am not sure how Miniconda could have such an adverse affect on our tests. So I am rebuilding libmesh/moose/tests again... just to be sure I didn't miss a step. If it fails again, I will yank Miniconda from my build, and try to re-build libmesh/moose/tests again... and see what happens.

Will keep you all informed.

Jason

To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/CAK06EuVVmKSBGTX5RD5gJdr1UZXeS1pjTDuNgHPLS6m0XSH7eg%40mail.gmail.com.

[MT]

I installed git using the following hyperlink:  http://git-scm.com/download/mac

I installed CMake using the following hyperlink:  http://www.cmake.org/files/v3.3/cmake-3.3.0-rc2-Darwin-x86_64.dmg

The trick with CMake was to drag the CMake file to the Applications Folder.  Once there, you can open it up and in the Tools menu it has an option "Install for Command Line Use".

I will try and run through the installation once more without Miniconda to see what happens.

I truly appreciate your help!

MT

[Miller, Jason M]

MT,

I would try to just move miniconda out of the way... no need to start _All_ over. Move it, then rebuild libmesh using the script. Proceed into the moose/tests directory and perform a `make cleanall` then attempt to rebuild moose and run tests.

I am in the process of rebuilding libMesh/moose with out miniconda present. I ran into the same errors after re-building everything twice now. So I am narrowing it down to miniconda.

Thanks for hanging in there!

Jason

To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/4cad3eb0-d7c4-41c1-b67e-0f9d02eb08f3%40googlegroups.com.

[Jean Francois Leon]

Jason
I tried that (opensuse 13.2 + gnome)
but for vtk here is what i got: ( missing dependency apparently)
 sudo -E zypper install  vtk
Loading repository data...
Reading installed packages...
Resolving package dependencies...

Problem: nothing provides libstdc++.so.6(GLIBCXX_3.4.21)(64bit) needed by vtk-6.2.0-42.1.x86_64
 Solution 1: do not install vtk-6.2.0-42.1.x86_64
 Solution 2: break vtk-6.2.0-42.1.x86_64 by ignoring some of its dependencies

[Miller, Jason M]

MT,

Removing Miniconda and then re-building everything starting with libMesh solved my issue of 575 failing tests on Snow Leopard. Don't know why it would affect the testing system so, but it seems to. With out Miniconda, you will most likely be with out Peacock functionality. To build all the dependencies Peacock requires by hand on Snow Leopard would be quite an undertaking.

Hope removing Miniconda will work for you!

Jason

[Cody Permann]

On another note, MOOSE should compile without vtk support. If that's not the case we need to fix it. The dependent pieces can just be skipped when support is turned off.

Cody

To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/CALgZTAcXJtSGoJ5J%2B16XVC%2BXywTy%2BPkQE1%3D%2BCenUwXsq%3DOOd6w%40mail.gmail.com.

[MT]

I ran the test without Miniconda which resulted in the following:

Ran 930 tests in 480.4 seconds
930 passed, 18 skipped, 0 pending, 0 failed

I am truly not sure why Miniconda caused it to fail before... but it seems to work without it.

This results in 2 questions:

Why were the 18 tests skipped (I attached the output)?

Without Peacock I will have to manually build or modify an input file and execute the application... is there a suggested software for post processing?

I am excited to dive into this software!

I truly appreciate you help and support getting me up and running!!

MT

[Andrew....@csiro.au]

MT, you don’t have to worry about “skipped” tests: they are irrelevant.

 

They are skipped because the developer wanted to test a particular thing which you don’t have (eg “using PETSC 3.5 req: < 3.4” means you are using PETSC3.5, but that particular test is only relevant if you are using PETSc version less than 3.4).  Or they are skipped because you are not running the test in the way the developer wanted (eg “METHOD!=dbg” means you ran the tests in optimized mode, but this test is only relevant for debug mode).

 

For postprocessing i use paraview.  i know a lot of other people do too.  It’s free, easy to install, easy to extend etc, and there are very few disadvantages with it.  You have to learn its features and how to drive it: that’ll take a day or two.

 

a

 

 

Ph: +61 7 3327 4497.  Fax: +61 7 3327 4666
Queensland Centre for Advanced Technologies
PO Box 883, Kenmore, Qld, 4069 
 

To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/f53ade98-84e5-4d6e-9257-66769ec8a7c7%40googlegroups.com.

[Miller, Jason M]

Glad its working for you!

And as for input file generation, I would take a look at the tests themselves for examples. You can also get all available input syntax by running:  <your application name> --dump  (in your case with moose/tests: `moost_test-opt --dump`)

Jason

To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/57596E6532C9F54F9DF09BCDFE7B1D3AA25FBE5A%40exmbx06-cdc.nexus.csiro.au.

[Miller, Jason M]

Jean,

Sorry for the delayed response... libstdc++ is a pretty straight forward available library. I am surprised you do not have that lib already present on your system (edit: you already do have it). Normally folks are needing to install an older version (5.0) for some such reason. Regardless, libstdc++6 should be a simple zypper install command.

(searched the internets)

There are some vague sites out there with folk mentioning the same missing library when in fact its sitting right there in /usr/lib... so why your package management system is having issues thinking its missing... sorry I got nothing!

Because you are doing this to get Peacock running, you may just want to install Miniconda as mentioned (but warned not to (: ) in the instructions.

Jason

To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/b9333d78-0ff2-48de-9522-61a10916a02a%40googlegroups.com.