Hi Drew, I just saw your reply by accident. It seems you did not copy me. Note that the submitter does not receive mails to n...@bugs.debian.org. Either use nnn-submitter@b.d.o. or add me directly.
> I checked with Nico upstream, cmake is not intended for building > pygalmesh for python module deployment. Ok, fine. I didn't want to suggest that the cmake build is better -- it is mainly about the discrepancy. After some debugging I found out why this happens: If you add "print (tmp_plugin, tmp_certainty)" in line 105 of /usr/bin/pybuild, you'll see that there is a tie between distutils and cmake and the first one tried wins. The order in which the plugins are tried is the file order in the directory /usr/share/dh-python/dhpython/build. In one chroot where I can reproduce the problem I see cmake first: $ ls -U1 /usr/share/dh-python/dhpython/build plugin_cmake.py plugin_custom.py base.py __init__.py plugin_distutils.py __pycache__ In another chroot where I can not reproduce the problem I see distutils first: $ ls -U1 /usr/share/dh-python/dhpython/build __init__.py plugin_distutils.py __pycache__ plugin_custom.py base.py plugin_cmake.py I guess the latter is also the case for the environment you used. I'll file a separate bug on dh-python. But since this is probably rather an implementation detail on their side, and for the being, I think a Build-Conflicts: cmake is in order for pygalmesh. Best regards, Joachim