We currently ignore coverage during up-front strategy generation (prep_args_kwargs_from_strategies in hypothesis). But we still collect coverage during dynamic data.draw calls, because HypofuzzProvider treats that as part of the test case.
Custom strategy code often has many branches, and I don't expect it to be useful feedback - guidance from test coverage inside strategies is already accounted for by the strategy domain itself.
We should consider ignoring dynamic draw coverage as well, if it can be done without too much overhead. A straightforward but likely too-expensive option is walking the stack for any hypothesis do_draw internals in each coverage event.
We currently ignore coverage during up-front strategy generation (
prep_args_kwargs_from_strategiesin hypothesis). But we still collect coverage during dynamicdata.drawcalls, becauseHypofuzzProvidertreats that as part of the test case.Custom strategy code often has many branches, and I don't expect it to be useful feedback - guidance from test coverage inside strategies is already accounted for by the strategy domain itself.
We should consider ignoring dynamic draw coverage as well, if it can be done without too much overhead. A straightforward but likely too-expensive option is walking the stack for any hypothesis
do_drawinternals in each coverage event.