Failing inflow_outflow.test

[trevilo]

The test inflow_outflow.test seems to fail sometimes when running via gitlab. The issues was first observed in regression testing on PR #223. It is not clear what is causing the failure, and so far it is not reproducible in a local build on the same machine (even though in theory everything is the same between the local build and the automated CI configuration).

The log from the failed test is pasted below:

1..5
ok 1 [inlet-outlet with argon minimal] check for input file inputs/argonMinimal.inflow_outflow.ini
PASS: inflow_outflow.test 1 [inlet-outlet with argon minimal] check for input file inputs/argonMinimal.inflow_outflow.ini
ok 2 [inlet-outlet with argon minimal] create a mesh file
PASS: inflow_outflow.test 2 [inlet-outlet with argon minimal] create a mesh file
ok 3 [inlet-outlet with argon minimal] create a restart file with a prescribed initial condition
PASS: inflow_outflow.test 3 [inlet-outlet with argon minimal] create a restart file with a prescribed initial condition
not ok 4 [inlet-outlet with argon minimal] run tps with input -> inputs/argonMinimal.inflow_outflow.ini
FAIL: inflow_outflow.test 4 [inlet-outlet with argon minimal] run tps with input -> inputs/argonMinimal.inflow_outflow.ini
# (in test file inflow_outflow.test, line 38)
#   `../src/tps -run $RUNFILE' failed with status 134
# 
# ------------------------------------
#   _______ _____   _____
#  |__   __|  __ \ / ____|
#     | |  | |__) | (___  
#     | |  |  ___/ \___ \ 
#     | |  | |     ____) | 
#     |_|  |_|    |_____/ 
# 
# TPS Version:  1.1 (dev)
# Git Version:  044f0f5
# MFEM Version: MFEM v4.4.1 (development)
# ------------------------------------
# 
# Options used:
#    --no-version
#    --runFile inputs/argonMinimal.inflow_outflow.ini
#    --no-debug
#    --no-visualization
# 
# Caching input file -> inputs/argonMinimal.inflow_outflow.ini
# 
# ---------------------------------
# MFEM Device configuration:
# Device configuration: hip,cpu
# Memory configuration: host-std,hip
# ---------------------------------
# 
# 
# TPS is using non-GPU-aware MPI.
# config.transportModel = constant
# name, input index, mixture index: E, 0, 1
# name, input index, mixture index: Ar, 1, 2
# name, input index, mixture index: Ar.+1, 2, 0
# 
# Inlet mass fraction of background species will not be used. 
# 
# Inlet mass fraction of electron will not be used. 
# Total # of mesh elements = 90
# number of elements on rank 0 = 90
# min elements/partition       = 90
# max elements/partition       = 90
# restart: 1
# HDF5 restart files mode: read
# Restarting from iteration = 0
# --> time = 0.000000e+00
# --> dt   = 1.000000e-05
# Restarting averages with 0
#  samplesReading in solutiond data from restart...
# --> Reading h5 path = /solution/density
# --> Reading h5 path = /solution/rho-u
# --> Reading h5 path = /solution/rho-v
# --> Reading h5 path = /solution/rho-E
# --> Reading h5 path = /solution/rho-Y_Ar.+1
# Reading in solutiond data from restart...
# --> Reading h5 path = /meanSolution/meanDens
# --> Reading h5 path = /meanSolution/mean-u
# --> Reading h5 path = /meanSolution/mean-v
# --> Reading h5 path = /meanSolution/mean-p
# --> Reading h5 path = /meanSolution/mean-YAr.+1
# Reading in solutiond data from restart...
# lt-tps: equation_of_state.cpp:592: double PerfectMixture::computeAmbipolarElectronNumberDensity(const double *) const: Assertion `n_e >= 0.0' failed.
# /var/tmp/oliver33/bats-run-249865-iVtpDl/bats.249896.src: line 37: 250100 Aborted                 ../src/tps -run $RUNFILE
not ok 5 [inlet-outlet with argon minimal] verify tps output with analytic solution -> inflow_outflow_output.ref.h5
FAIL: inflow_outflow.test 5 [inlet-outlet with argon minimal] verify tps output with analytic solution -> inflow_outflow_output.ref.h5
# (in test file inflow_outflow.test, line 48)
#   `h5diff --relative=$TOL $SOLN_FILE $REF_FILE /solution/rho-Y_Ar.+1' failed
# dataset: </solution/rho-Y_Ar.+1> and </solution/rho-Y_Ar.+1>
# 90 differences found
ERROR: inflow_outflow.test - exited with status 1

It indicates a failing assert on the electron number density.

[trevilo]

In addition to not being able to reproduce locally, the tests passed when run after PR #223 was merged. So... very confusing.