Open jku opened 2 months ago
I suppose it is possible that some expected client failure actually happens differently without metadata.py (like because the repository fails to process something that it did manage to process with metadata.py)... will need to compare the repository dumps
The code in metadata.py is either never executed (this is 70% of it) or does not affect the test results in any way: I propose we remove the unused code since it gives the impression there is something clever being done but it's mostly not even used.
We can re-add code like this when we have a test that demonstrably needs it.