I would think the python-gnucash dependency issue is a packaging bug. One could mark python-gnucash as a mandatory dependency. That may cause circular dependencies though because python-gnucash should depend on gnucash.
The other option is alter the package contents slightly. (Caveat: I'm not a packager so I may be talking nonsense)
From gnucash' point of view the python binding are really optional. It will not try to load any python script if the module libgncmod_python.so is not found. So I think that if this module is moved to the python-gnucash package, python-gnucash can really be a true optional module.
There may be one issue: if python-gnucash is an architecture independent package, libgncmod_python.so can't just moved into that package because that would make it architecture dependent. If that's the case, I'll have to pass the issue to real packagers...
I would think the python-gnucash dependency issue is a packaging bug. One could mark python-gnucash as a mandatory dependency. That may cause circular dependencies though because python-gnucash should depend on gnucash.
The other option is alter the package contents slightly. (Caveat: I'm not a packager so I may be talking nonsense)
From gnucash' point of view the python binding are really optional. It will not try to load any python script if the module libgncmod_python.so is not found. So I think that if this module is moved to the python-gnucash package, python-gnucash can really be a true optional module.
There may be one issue: if python-gnucash is an architecture independent package, libgncmod_python.so can't just moved into that package because that would make it architecture dependent. If that's the case, I'll have to pass the issue to real packagers...