David Neto

Checking C++ std::numeric_limits * lowest: Returns the lowest finite value representable by the numeric type T, that is, a finite value x such that there is no other finite value...

The proposal was posted live in the meeting and we didn't have the normal preparation period. I've found some problems with how it interacts with parsing. The bitcast precedent works...

To be concrete about it, consider: ``` type myfloat = f32; const a = 1; const b = 2; const blah = b` to see it isn't // like the...

Decided at F2F 2019-05-16: [Minutes doc](https://docs.google.com/document/d/1HR9vPk48-lhzHAsyF57QzjKMmi0gQHfSXMmyqZuFWkc/edit#) Implementation behaves as if the shift count is masked with (bitwidth-1). (e.g. if the bit width is 32 then mask the shift count by...

Right, thanks! The idea is that we already have operations on reference-types (think l-value in C/C++): * `s.field` gets from an l-value struct `s` to an l-value of its `field`...

Note, this proposal still plays very nicely with a future with type derivation for var and const declarations. That was something that the original #1368 did not have. So that's...

I forgot to mention another difference between references and pointers: * The LHS of an assignment statement must be a reference.

Discussed in meeting of [2021-03-23](https://docs.google.com/document/d/19inX49uQBMkWx1lSuPkFWJ8fGRJUGMNXmkUj5TJWnwY/edit#) * Consensus on accepting this proposal. * Strongly consider adding -> as convenience / syntactic sugar. * Proposed by Apple, acceptable to Google (but prioritized...

Thanks for catching this. Indeed it's a conflict. My gut feeling is is `enable f16` should be allowed: extension names should be as free as possible. This could be done...

`IDENT` in the grammar is not `identifier` in the spec. `IDENT` in the grammar is defined by a regular expression, and that doesn't exclude keywords. Also, if we went down...