Open cdonovick opened 1 year ago
Due to some sort of caching issue I can't test BrokenProtocol
and FixedProtocol
at the same time. Not sure how to resolve.
Is the cache issue at the magma/python level compile? Or the verilator test?
Is the cache issue at the magma/python level compile? Or the verilator test?
Sorry I should have provided more detail. If I add FixedProtocol
to the list of BoxT
I get an attribute error about BrokenProtocol
having no debug_name
in the FixedProtocol
cases (this is the error I get in the BrokenProtocol
cases). If I run only FixedProtocol
it has no errors. If I set BoxT = [FixedProtocol, BrokenProtocol]
I get no errors. So some sort of Python level caching issue.
I think a broader fix for this issue is to have fault
unpack protocol values at the entrypoint to any fault code (i.e. the fault internals just operate on the underlying magma values). I added the required unpacking for these tests, but I suspect there are (many) other places we may need to the get_magma_value
logic. However, this is likely much simpler to do at the API boundary rather than finding all the places internally where we need to call get_magma_value
Related issue #327