On 19 July 2018 at 20:18, Jonathan Wakely <jwak...@redhat.com> wrote: > This removes some seemingly redundant conditions from a few traits. If > __is_trivially_assignable correctly checks the assignable condition as > well as triviality, then we don't need is_assignable explicitly. Does > anybody see a problem with that
It should work; if it doesn't, that's a bug in the compiler. Both is_constructible and is_assignable and trivial variants thereof go to the same "is_xible" code path, so it should be fair for the library to expect that it all works. In case it doesn't, that's something that I will fix in the front-end. > I added some extra tests for cases that had been problematic with > __is_trivially_constructible. That seems reasonable; some of those tests are now duplicated on the compiler side, but I think that's fine.
