Issue when importing podio from python in debug builds: AttributeError: <namespace cppyy.gbl.podio.version at 0x7a0dec0> has no attribute 'build_version'.

[jmcarcell]

When importing podio in debug builds the following error appears:

cling JIT session error: Failed to materialize symbols: { (main, { _ZN5podio7versionL13build_versionE }) }
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/cvmfs/sw-nightlies.hsf.org/key4hep/releases/2024-11-18/x86_64-almalinux9-gcc14.2.0-dbg/podio/5afdefdd233bd6b53f51d38bceffae600ae41409_develop-o5edi4/lib/python3.11/site-packages/podio/__init__.py", line 14, in <module>
    from . import root_io, reading, version
  File "/cvmfs/sw-nightlies.hsf.org/key4hep/releases/2024-11-18/x86_64-almalinux9-gcc14.2.0-dbg/podio/5afdefdd233bd6b53f51d38bceffae600ae41409_develop-o5edi4/lib/python3.11/site-packages/podio/version.py", line 41, in <module>
    build_version = podio.version.build_version
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: <namespace cppyy.gbl.podio.version at 0x7a0dec0> has no attribute 'build_version'. Full details:
  type object 'version' has no attribute 'build_version'
  'podio::version::build_version' is not a known C++ class. Did you mean: 'decode_version'?

This can be consistently reproduced with the current nightlies both on Ubuntu 22.04 and Alma 9. Locally with the same ROOT version and the same build options I can not reproduce this. I know of two workarounds:

1. Remove FCCAnalyses and EDM4hep from LD_LIBRARY_PATH.
2. Remove static constexpr from this line https://github.com/AIDASoft/podio/blob/master/podioVersion.in.h#L72

I think it has to do with ROOT not being happy with the build_version symbol being in the edm4hep libraries that it finds through LD_LIBRARY_PATH, something that in the opt builds gets optimized out and it can't be found in the edm4hep libraries.

[andresailer]

We get the same in dev4 (Root 6.32-patches) lcgstacks https://lcgapp-services.cern.ch/cdash3/tests/18944514

but not in dev3 (Root master) https://lcgapp-services.cern.ch/cdash3/tests/19035061#

[jmcarcell]

I think the plan is going to be to wait to see if this happens the next time ROOT is updated.