Laurence Tratt

Via Ted Unangst, Darwin [also skips negative fds in poll](https://github.com/apple/darwin-xnu/blob/main/bsd/kern/sys_generic.c#L1886) and, interestingly, has this comment in the code: ``` /* per spec, ignore fd values below zero */ ```

@SteveLauC If one can avoid the extra `Vec` that would be cool. FWIW I have (but currently untested while I change the program that would call this!) locally: ```diff diff...

Thank you @SteveLauC -- much appreciated!

Oops, I missed this one entirely! I think I'm broadly in favour of this, though we'd have to define what happens if a user specifies `YaccGrammarKind` *and* the grammar specifices...

Good question, and I don't immediately have a good suggestion. @ratmice might?

> There are also some questions if we are making the id public if it should also have an idxnewtype associated with it I suppose? What do we expect people...

Ah, I see, so I *think* we're saying: 1. Ideally we want to return `Option`. 2. But that would require `clone`ing a `StartState` which commits us to an ugly API....

Hello Simon, I'm keen to get a new release of grmtools out of the door soon, as we've piled up lots of useful changes since the last release 8 months...

Sorry! By "opaque type" we just mean `struct S(SomeHiddenType)` where users can't access `SomeHiddenType` directly but can still e.g. compare two `S`s for equality, or hash `S`s etc. In this...

@SMartinScottLogic Please squash (you'll need to rebase against master and resolve whatever the conflict(s) is/are).