On Wed, Mar 17, 2021 at 04:45:00PM -0400, Jason Merrill wrote: > On 3/8/21 7:52 PM, Marek Polacek wrote: > > Yeah; it's worked pretty well for classes after we've dealt with the > > initial fallout. I've factored the check into a new function, and also > > extended it with the case where we'd create a FLOAT_EXPR. I don't know > > if that covers all non-trivial conversions. > > By non-trivial I was thinking pretty much any non-identity conversion. But > your patch is probably safer for trunk/10. OK.
Yea, I'd leave it as-is for now. > > Do we want to assert that FLOAT_EXPR never makes its way into tsubst now? > > Sure. Ok, I'll add the assert if it passes testing. Thanks, Marek