while writing some other bits and pieces, I had a use case for checking the warnings omitted, RunResult has a assert_outcomes() that doesn't quite offer warnings= yet the information is already available in there, I suspect there is a good reason why we don't have assert_outcomes(warnings=...) so I propose some additional capabilities on RunResult to handle warnings in isolation.

With assert_outcomes() the full dict comparison may get a bit intrusive as far as warning capture is concerned.

something simple like:

result = pytester.runpytest(...)
result.assert_warnings(count=1)

Thoughts?