https://github.com/ojhunt commented:
In general this looks fine but I'd like @shafik's feedback as well.
My only real question is the reason for not using scoped enums? This is a
genuine question outside of just this PR, as enum classes trade off a bunch of
convenience in exchange for there tighter type constraints.
I'd certainly prefer enum class as it makes the enumerations read better and
avoids accidental bool usage, but I get that it's trade that off against a lot
of verbosity and the lack of some basic `if(scopedenumeration)`, etc syntax.
I don't consider this a "must change" consideration, but for instance:
```cpp
enum class MemberInitKind {Default, NonDefault}
```
gives you `MemberInitKind::Default` vs `DefaultMemberInit` which isn't hugely
shorter, and seems subjectively better to me, however I recognize it means that
in other cases this same pattern causes a lot of verbosity in things like
`allowsNRVO()` where you end up with `LocAndNRVO.NRVOMode == NRVOKind::Allowed`
under the same scheme.
So again I don't consider this a blocker to the PR, I just want to know if
scoped enums were considered and dismissed (partially because I want to do
something about some of the annoyances of scoped enums).
https://github.com/llvm/llvm-project/pull/185335
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
