Over in #1191, I found that there was a issue with the use of patching in tests of the static settings in model init and _setup, which was masking some missing init vars.
This is being done because the tests use compiled sets of test data that include variables created by init and update, so the static settings were being patched to avoid raising configuration errors. That's a problem because it breaks the explicit variable checking - for example, in the abiotic model it was masking incomplete required vars lists - the variable might not be in the required list, but is in the data anyway, so the test runs.
Over in #1191, I found that there was a issue with the use of patching in tests of the static settings in model init and _setup, which was masking some missing init vars.
This is being done because the tests use compiled sets of test data that include variables created by init and update, so the static settings were being patched to avoid raising configuration errors. That's a problem because it breaks the explicit variable checking - for example, in the abiotic model it was masking incomplete required vars lists - the variable might not be in the required list, but is in the data anyway, so the test runs.