Open garlick opened 10 months ago
One note: if I run flux make check
, the tests work, so perhaps something environmental with the unit tests.
I wonder if the flux-accounting Python bindings is doing something incorrectly with import
ing the flux
Python module? Sorry, I feel like I am no help here 😩 FWIW, I think the Makefile.am
for flux-accounting is the same as flux-sched's fluxion Makefile.am
:
SUBDIRS = accounting
install-data-hook:
$(AM_V_at)echo Linking python modules in non-standard location... && \
$(INSTALL) -d -m 0755 "$(DESTDIR)$(fluxpylinkdir)" && \
target=$(fluxpydir) && \
f=$${target##*/} && \
cd "$(DESTDIR)$(fluxpylinkdir)" && \
rm -f $$f && \
$(LN_S) $$target .
uninstall-local:
$(AM_V_at)target=$(fluxpydir) && f=$${target##*/} && \
echo "Removing $(fluxpylinkdir)/$$f" && \
rm -rf $(fluxpylinkdir)/$$f
I was thinking maybe the python environment for unit tests would be the right place to start looking:
AM_TESTS_ENVIRONMENT = \
export PYTHONPATH="$(abs_top_srcdir)/src/bindings/python/:$(abs_top_srcdir)/t/pycotap:$(PYTHON_SITE_PKG):$$PYTHONPATH";\
export PYTHON="${PYTHON}";
But that looks ok to me. I must be missing something.
Problem: as noted here, one of the unit tests is failing for me:
This is a debian bullseye system with flux-core installed as a system package.