Due to the layout of the loop creating the related_schema isntances, all
but the last related_include were lost. They are thus now collected in
the inital step where all include_paths are evaluated, so that the
schema creation code can use the colelcted sub-includes later.
It is still not perfect, because the related_schema is created multiple
times and later discarded in some cases, but that's a minor issue and
was there before as well. The important thing is that the last schema
created is complete now.
Coverage increased (+0.009%) to 95.048% when pulling a000bc776098da313f9cf5c5b2e41e35e59e7fe0 on Natureshadow:issue-56 into 9566f6f9ae371d0e348c8a8804d3b27dcb4dc2c6 on miLibris:master.
Due to the layout of the loop creating the related_schema isntances, all but the last related_include were lost. They are thus now collected in the inital step where all include_paths are evaluated, so that the schema creation code can use the colelcted sub-includes later.
It is still not perfect, because the related_schema is created multiple times and later discarded in some cases, but that's a minor issue and was there before as well. The important thing is that the last schema created is complete now.