Rust 1.74 #16416
Rust 1.74 #16416
Conversation
To 8c0ed39e4797dafc6e5e3486dc4fcebec4543623
Bump up to 0.2.90 instead of 0.2.87
It seems that clippy is not run in this crate. We should activate it later.
Output given when simply executing `cargo build --release` in the directory
Use 25c111e2a35fb9e953e3b6378a3982d85bc6bac3, commit removing the unused gate types related to Keccak, instead of propagating the changes in the whole codebase.
Bringing back ocaml_str, which is used in gen_scalars.ml
The method is used in src/lib/pickles to generate scalars equations, and the variables are used from the cache to generate the OCaml code.
Moving secp256k1 into dev-dependencies to avoid compiling it when generating wasm library
Removing secp256k1 as a dependency as it was simply used in testing in proof-systems. It has been previously added by mistake.
Removing secp256k1 as a dependency as it was simply used in testing in proof-systems. It has been previously added by mistake.
To have the same version than proof-systems
Use the commit than downgrades wasm-bindgen to 0.2.87
|
!ci-nightly-me |
| // simultaneously, and | ||
| // * the underlying pointer is in the correct state to be | ||
| // mutable, since we can call [`get_mut_unchecked`] in | ||
| // nightly, or can call [`get_mut`] and unwrap if this is | ||
| // the only live reference. | ||
| &mut *(((&*self.0) as *const Self::Target) as *mut Self::Target) |
There was a problem hiding this comment.
1.74 throws a compiler error here. There might be some other way to get it to stop complaining, but I think this is equivalent without the illegal cast
| @@ -1,7 +1,7 @@ | |||
| macro_rules! impl_caml_pointer { | |||
| ($name: ident => $typ: ty) => { | |||
| #[derive(std::fmt::Debug, Clone, ::ocaml_gen::CustomType)] | |||
| pub struct $name(pub ::std::rc::Rc<$typ>); | |||
| pub struct $name(pub ::std::rc::Rc<std::cell::UnsafeCell<$typ>>); | |||
There was a problem hiding this comment.
Note as it is important for the Caml binding: UnsafeCell is implemented as:
#[lang = "unsafe_cell"]
#[stable(feature = "rust1", since = "1.0.0")]
#[repr(transparent)]
#[rustc_pub_transparent]
pub struct UnsafeCell<T: ?Sized> {
value: T,
}
There was a problem hiding this comment.
I guess we're adding a new indirection? We might lose some performances?
There was a problem hiding this comment.
Note as it is important for the Caml binding: UnsafeCell is implemented as:
The nightly tests were passing with this change, does this mean that that the ocaml bindings lib was able to resolve this on its own or is it just coincidence because maybe the runtime rep is the same
There was a problem hiding this comment.
I guess we're adding a new indirection? We might lose some performances?
My intuition is that this UnsafeCell should be optimized away. We can discuss
| // simultaneously, and | ||
| // * the underlying pointer is in the correct state to be | ||
| // mutable, since we can call [`get_mut_unchecked`] in | ||
| // nightly, or can call [`get_mut`] and unwrap if this is | ||
| // the only live reference. | ||
| &mut *(((&*self.0) as *const Self::Target) as *mut Self::Target) | ||
| &mut *self.0.get() |
There was a problem hiding this comment.
Note: implementation of get.
#[inline(always)]
#[stable(feature = "rust1", since = "1.0.0")]
#[rustc_const_stable(feature = "const_unsafecell_get", since = "1.32.0")]
#[rustc_never_returns_null_ptr]
pub const fn get(&self) -> *mut T {
// We can just cast the pointer from `UnsafeCell<T>` to `T` because of
// #[repr(transparent)]. This exploits std's special status, there is
// no guarantee for user code that this will work in future versions of the compiler!
self as *const UnsafeCell<T> as *const T as *mut T
}
There was a problem hiding this comment.
Nit: cargo fmt reformats this line. It seems we do not run in the Mina CI clippy nor fmt.
There was a problem hiding this comment.
Note: implementation of get.
This makes me think someone was trying to get around using UnsafeCell when they wrote this the first time. I would hope that this would be optimized away
|
LGTM so far, modulo comments. |
|
Feel free to cherry-picked some changes from #16411 |
martyall
force-pushed
the
martin/rust-1.74
branch
from
33e81df to
308fa27
Compare
dannywillems
force-pushed
the
dw/use-latest-master-proof-systems
branch
2 times, most recently
from
ed00f55 to
f6f23c1
Compare
Depends on #16405 and proof-systems-2868 and proof-systems-2875
This PR updates the rust compiler to
1.74. It is a stepping stone in updating thewasm-packandwasm-bindgen-clidependencies, but it is also good housekeeping to not fall too far behind.