Fix recursion limits with next solver

CHANGELOG.md

@@ -82,6 +82,7 @@ Bottom level categories:
 - Zero-initialize padding (if any) at the end of a buffer allocation. This was application-visible in rare cases on Vulkan when a shader read beyond the valid range of a vertex buffer. By @andyleiserson in [#9791](https://github.com/gfx-rs/wgpu/pull/9791).
 - Fix required immediate slots calculation and remove `naga::valid::FunctionInfo::immediate_slots_used`. By @beicause in [#9725](https://github.com/gfx-rs/wgpu/pull/9725).
 - Fix `PendingSubmission` releasing its lock guards out of stacking order, which tripped `--cfg wgpu_validate_locks` on any submission. By @AdrianEddy in [#9960](https://github.com/gfx-rs/wgpu/pull/9960).
+- Explicitly implement `Send` and `Sync` in `wgpu-core` to reduce the risk of `recursion_limit` overflow. By @nazar-pc in [#9953](https://github.com/gfx-rs/wgpu/pull/9953)
 
 #### naga
 
@@ -1481,7 +1482,7 @@ By @wumpf in [#8282](https://github.com/gfx-rs/wgpu/pull/8282), [#8285](https://
 - naga now requires that no type be larger than 1 GB. This limit may be lowered in the future; feedback on an appropriate value for the limit is welcome. By @andyleiserson in [#7950](https://github.com/gfx-rs/wgpu/pull/7950).
 - If the shader source contains control characters, naga now replaces them with U+FFFD ("replacement character") in diagnostic output. By @andyleiserson in [#8049](https://github.com/gfx-rs/wgpu/pull/8049).
 - Add f16 IO polyfill on Vulkan backend to enable SHADER_F16 use without requiring `storageInputOutput16`. By @cryvosh in [#7884](https://github.com/gfx-rs/wgpu/pull/7884).
-- For custom Naga backend authors: `naga::proc::Namer` now accepts reserved keywords using two new dedicated types, `proc::{KeywordSet, CaseInsensitiveKeywordSet}`. By @kpreid in [#8136](https://github.com/gfx-rs/wgpu/pull/8136).
+- For custom Naga backend authors: `naga::proc::Namer` now accepts reserved keywords using two new dedicated types, `proc::{KeywordSet, CaseInsensitiveKeywordSet}`. By @kpreid in [#8136](https://github.com/gfx-rs/wgpu/pull/8136).
 - **BREAKING**: Previously the WGSL storage-texture format `rg11b10float` was incorrectly accepted and generated by naga, but now only accepts the the correct name `rg11b10ufloat` instead. By @ErikWDev in [#8219](https://github.com/gfx-rs/wgpu/pull/8219).
 - The [`source()`](https://doc.rust-lang.org/std/error/trait.Error.html#method.source) method of `ShaderError` no longer reports the error as its own source. By @andyleiserson in [#8258](https://github.com/gfx-rs/wgpu/pull/8258).
 - naga correctly ingests SPIR-V that use descriptor runtime indexing, which in turn is correctly converted into WGSLs binding array. By @hasenbanck in [8256](https://github.com/gfx-rs/wgpu/pull/8256).

wgpu-core/src/global.rs

@@ -287,6 +287,47 @@ impl Global {
     }
 }
 
+/// Implement [`Send`] + [`Sync`] for [`Global`], and check that all of its fields are.
+///
+/// This is identical to the “auto trait” implementation that Rust would provide, except that
+/// it is eager rather than lazy: its requirements are checked now (in
+/// `_global_fields_are_send_sync`), when this crate is compiled, rather than whenever a dependent
+/// wants to know whether `Global: Send` holds.
+///
+/// This improves compilation performance and avoids a risk of dependents running into the default
+/// [`recursion_limit`] when checking types containing [`Global`]. This risk will become greater
+/// when Rust’s “next solver” is stabilized.
+///
+/// [`recursion_limit`]: https://doc.rust-lang.org/reference/attributes/limits.html#the-recursion_limit-attribute
+#[cfg(send_sync)]
+mod global_send_sync {
+    use super::Global;
+
+    // SAFETY: Bounds checked below
+    unsafe impl Send for Global {}
+
+    // SAFETY: Bounds checked below
+    unsafe impl Sync for Global {}
+
+    /// This function will fail to compile if any field is not `Send + Sync`, or if a new field is
+    /// added to `Global`.
+    ///
+    /// This technique is modeled after the macro library <https://crates.io/crates/non_structural_derive>;
+    /// however, we only need it once, so it’s cheaper to write out the same code the macro would generate.
+    fn _global_fields_are_send_sync(global: &Global) {
+        fn _check_bound<T: Send + Sync>(_: &T) {}
+
+        let Global {
+            surfaces,
+            hub,
+            instance,
+        } = global;
+        _check_bound(surfaces);
+        _check_bound(hub);
+        _check_bound(instance);
+    }
+}
+
 impl fmt::Debug for Global {
     fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
         f.debug_struct("Global").finish()
@@ -299,9 +340,3 @@ impl Drop for Global {
         resource_log!("Global::drop");
     }
 }
-
-#[cfg(send_sync)]
-fn _test_send_sync(global: &Global) {
-    fn test_internal<T: Send + Sync>(_: T) {}
-    test_internal(global)
-}

wgpu-core/src/lib.rs

@@ -7,8 +7,6 @@
 //!
 
 #![no_std]
-// `-Znext-solver` requires deeper recursion limits (at least for now) to prove Send/Sync
-#![recursion_limit = "256"]
 // When we have no backends, we end up with a lot of dead or otherwise unreachable code.
 #![cfg_attr(
     all(

wgpu/src/lib.rs

@@ -216,8 +216,6 @@
 //!
 
 #![no_std]
-// `-Znext-solver` requires deeper recursion limits (at least for now) to prove Send/Sync
-#![recursion_limit = "256"]
 #![cfg_attr(docsrs, feature(doc_cfg))]
 #![doc(html_logo_url = "https://raw.githubusercontent.com/gfx-rs/wgpu/trunk/logo.png")]
 #![warn(