Obscure error with incorrect keyword argument defaults

[c42f]

If a keyword argument has a declared type, but the default value is of a different type, the error message is confusing:

julia> f(; a::Float64=1) = a
f (generic function with 1 method)

julia> f()
ERROR: MethodError: `__f#0__` has no method matching __f#0__(::Int64)

It would be nice to have a more comprehensible error message here. (Could this error even be caught during type inference? It seems like all the information should be there.)

[c42f]

Side note - I tried using code_typed to understand what is produced here when a is supplied vs defaulted, but I don't see a way to specify keyword arguments to code_typed. Is this possible?

[mauro3]

Struggling with this error message over in ODE.jl: SciML/ODE.jl#49 (comment)

[JeffBezanson]

Very similar to #16490