lang-team VS 708

The Rust team did a deep dive on the bug in 2020, which has some more details that might be helpful to understanding what's going on: https://github.com/rust-lang/lang-team/blob/master/design-me....

I haven't been following this closely, so I looked it up and it looks like that's not going to happen for the foreseeable future unfortunately:

https://github.com/rust-lang/lang-team/issues/122

Kind of a shame, but wrapper types work well enough that I understand. It does look like if there was someone with enough resources to make it happen that they'd be receptive to it.

Did you read the link the original comment posted? I think that explains the idea rather well https://github.com/rust-lang/lang-team/issues/122

A quick search pulled up this as a likely candidate for most recent discussion of it but it goes back at least to 2016 with this RFC.

There have been at least one proposal and RFC in the past that seem to be deferred or closed due to bandwidth issues.

That's what Rust does, and it's considered a problem (that the devs are regrettably unable to reasonably solve) rather than a good thing.

Lang had some conversations about this: https://github.com/rust-lang/lang-team/issues/139

The design of Rust panics unconditionally aborts the program if you panic while unwinding, and some people even want to abort if you panic in Drop.

> Better interoperability

AFAIK, the examples you give all target a basic C ABI [0] or can be made to target the same ABI. In Rust, it means targeting wasm32-unknown-emscripten

The Rust team is also working on a "WASM ABI"[1] which would be useful in taking advantage of stuff like multi-value returns, and other compilers could just choose to target that. More likely, the C ABI on WASM will be updated to account for missing features, and that'll be the standard for interoperability in the WASM ecosystem.

[0]: https://github.com/WebAssembly/tool-conventions/blob/main/Ba...

[1]: https://github.com/rust-lang/lang-team/blob/master/design-me...

Herb's parameter passing paper describes five parameter semantics categories (in, inout, out, move, and forward) to replace C++'s parameter passing styles (value, reference, const reference, r-value reference, and forwarding reference). And it poses the question of how parameter passing semantics can reasonably be made visible at the call-site.

Another design mistake, imho, is that Rust does not have Herb Sutter’s parameter passing style from his 708 paper. For example, out parameters instantly remove almost all use-cases for MaybeUninit, adding strong compiler guarantees and requiring zero unsafe code.

In my C++ code, I've copied Rust and expressed fallible initialization by returning a std::optional from a static method (which the article failed to mention). The problem is that (like Rust) you lose placement initialization, and (unlike C++ or Rust) you can't initialize private fields using aggregate initialization or initializer lists, and must write a passthrough constructor (which can't even be private because it breaks make_unique).

https://github.com/hsutter/708 is a C++ proposal which unifies placement constructors and writable out-parameters ("definite first use"). I don't think it makes placement initialization fallible, but I'm not sure.

As noted elsewhere - we need in/out/forward.

There is a proposal by Herb Sutter (708: Parameter passing -> guaranteed unified initialization and unified value-setting) that want to do just that.

Having said this, I can't wait for the dream to become reality

I am waiting for this to become reality, and C++ language to suddenly lose 98% of its annoyingness!

Oh! That would be so nice. I mean, there is a proposal in the works, but I really doubt it will be adopted as it is.