EXODIFF Test Failures and "EXPECTED OUTPUT MISSING" on MOOSE Commit b9efdf48dd: Installation Issue? #31188
Replies: 2 comments 2 replies
-
|
Hello
Is fine. We have seen this off failure creep up when we change the list of default created objects. No action to take here we ll get to it.
These two are odd. The images are not being read correctly, they are read as empty. However, unless you are planning to create meshes from PNG pictures, as the phase field team did, you don't need this so you could consider ignoring it. To your questions:
No we test across several linux and macos without failures. They are not expected, but unless we test your exact platform and dependencies we also cannot guarantee you won't get any failure.
You should not wait to pass all tests before learning moose. The vast majority of the tests passed. Your version of moose is functional at this time. In fact, I think once you get comfortable moose you will be able to fix those tests.
Yes your moose is as recent as it can be.
Thanks for the snapshots. A log is preferred but this works. Unfortunately nothing obvious. Some package must not be behaving as expected for reading these figures. If you need this feature specifically we can look more into it |
Beta Was this translation helpful? Give feedback.
-
|
Hi, Thank you so much for the clarification and reassurance — that really helped! I’ve now started using MOOSE without worrying about the minor test failures. Since I’m not planning to use image-based meshing for now, I’ll safely ignore the image_function-related errors. Some Follow-up Clarifications:
XDR-related errors (possibly from libtirpc and related libraries) LD_LIBRARY_PATH conflicts, where Conda libraries and system libraries clashed Compiler mismatches, due to mixing system GCC and Conda compilers unintentionally After troubleshooting, I switched to installing and running MOOSE outside Conda, in the regular system shell, which worked much more cleanly. Given that I plan to run MOOSE later on Isambard AI (an HPC platform), do you recommend continuing with this “normal terminal” setup (outside Conda)? Thanks again for your support and for maintaining such a great ecosystem! Best, |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Check these boxes if you have followed the posting rules.
Issue or question about MOOSE
Hi MOOSE developers and community,
I recently built MOOSE from commit b9efdf4 (dated 2025-07-31) on my Ubuntu system, following the official build guidelines. basically I followed what is written here "https://mooseframework.inl.gov/getting_started/installation/gcc_install_moose.html". I cloned moose and just allowed moose to install and manage petsc and libmesh and wasp etc etc.
The build completes successfully, but when I run the test suite, I encounter multiple test failures: All in all I had 13 failures(12 of them are exodiff failures from image function tests, and 1 on expected output missing).
**1. EXODIFF Failures on functions/image_function tests
Examples of failure messages from exodiff show large relative differences in nodal variable u at node 311:
exodiff: Files are different
Nodal variables:
u rel diff: 5.7223e+04 ~ 0.0 = 1.0 (node 311)**
****2. minimal_app.minimal test failure with "EXPECTED OUTPUT MISSING"
./run_tests --re=minimal_app.minimal
Unable to match the following pattern against the program's output:
DefaultNonlinearConvergence\nFEProblem\nGeneratedMesh
Tester failed, reason: EXPECTED OUTPUT MISSING
The test output shows the listed objects but does not match the expected pattern.**
System and Build Details:
MOOSE commit: b9efdf4 (2025-07-31)
PETSc version: 3.23.0
LibMesh version: f0a69d16b43414eba6d1c1bfe8482ccd6bda220b
VTK version: 9.1 (system installed)
OS: Ubuntu 24.04.2 LTS
Build: Release mode with standard dependencies, no conda environment used during build or test
Diagnostics (scripts/diagnostics.sh) report that the compiler environment variables (CC, CXX, FC, etc.) are not set, causing the check to fail.(strange!!!). I am running all commands outside Conda, in the normal terminal environment, though my PATH shows a Conda condabin at the front. Also, Python version is 3.12.3 and Python sanity checks pass.
Could the missing compiler environment variables be causing the test failures? Should I explicitly set CC, CXX, FC, etc., before building? Any advice on resolving the EXODIFF issues would also be appreciated.
$ which cc
which gcc
which mpicc
which c++
which g++
which mpicxx
which fc
which gfortran
which mpif90
/usr/bin/cc
/usr/bin/gcc
/usr/bin/mpicc
/usr/bin/c++
/usr/bin/g++
/usr/bin/mpicxx
/usr/bin/gfortran
/usr/bin/mpif90
So the compilers are okay, and i manually tested they are working fine in terminal. But if i echo $CC, $CCX, $FC they return zero, meaning they are not set! That could be the diagnostics issue, but still does not explain exodiff failure and expected output missing errors!
Questions:
Thanks in advance for your help!
(Optional) code in question / simulation log / errors
Encountering Errors? Please include diagnostic output
Beta Was this translation helpful? Give feedback.
All reactions