Make TypeId const comparable #142789

oli-obk · 2025-06-20T16:13:56Z

This should unblock stabilizing const TypeId::of and allow us to progress into any possible future we want to take TypeId to.

To achieve that TypeId now contains 16 / size_of<usize>() pointers which each are actually just size_of<usize>() bytes of the stable hash. At compile-time the first of these pointers cannot be dereferenced or otherwise inspected (at present doing so might ICE the compiler). Preventing inspection of this data allows us to refactor TypeId to any other scheme in the future without breaking anyone who was tempted to transmute TypeId to obtain the hash at compile-time.

cc @eddyb for their previous work on #95845 (which we still can do in the future if we want to get rid of the hash as the final thing that declares two TypeIds as equal).

Tracking Issue for const fn type_id #77125

r? @RalfJung

rustbot · 2025-06-20T16:14:02Z

Some changes occurred to the intrinsics. Make sure the CTFE / Miri interpreter
gets adapted for the changes, if necessary.

cc @rust-lang/miri, @RalfJung, @oli-obk, @lcnr

Some changes occurred in compiler/rustc_codegen_cranelift

cc @bjorn3

Some changes occurred to the CTFE / Miri interpreter

cc @rust-lang/miri, @RalfJung, @oli-obk, @lcnr

Some changes occurred to the CTFE / Miri interpreter

cc @rust-lang/miri

Some changes occurred to the CTFE machinery

cc @RalfJung, @oli-obk, @lcnr

Some changes occurred in compiler/rustc_codegen_gcc

cc @antoyo, @GuillaumeGomez

library/core/src/intrinsics/mod.rs

library/core/src/any.rs

RalfJung · 2025-06-21T10:37:29Z

It will be a while until I have the capacity to review a PR of this scale.

Meanwhile, could you say a bit more about the architecture of the change? It seems you want for the "new kind of allocation" approach, but it's not clear from the PR description how exactly that shows up in TypeId.

Also, I am definitely not comfortable landing this by myself, I can only review the const-eval parts. Changing the representation of TypeId has ramifications well beyond that that I do not feel qualified to evaluate -- I think an MCP would be justified.

rustbot · 2025-06-21T11:17:11Z

Some changes occurred in compiler/rustc_codegen_ssa

cc @WaffleLapkin

oli-obk · 2025-06-21T11:18:41Z

Well, I got private feedback yesterday that instead of encoding a 16 byte value as an 8 byte pointer to the 16 byte value and an 8 byte hash, I should just do the thing where we split up type id internally into pointer sized chunks and codegen will make a hash out of it again.

TLDR: no changes to runtime type id anymore in the latest revision of this PR. Only compile-time type id is now a bit funny

oli-obk · 2025-06-21T11:22:11Z

It will be a while until I have the capacity to review a PR of this scale.

I'm splitting unrelated parts out, so the high level feedback is already useful and I'll look for libs and codegen ppl to review the appropriate parts

compiler/rustc_codegen_gcc/src/common.rs

compiler/rustc_codegen_cranelift/src/constant.rs

Make `PartialEq` a `const_trait` r? `@fee1-dead` or `@compiler-errors` something generally useful but also required for rust-lang#142789

Make `PartialEq` a `const_trait` r? ``@fee1-dead`` or ``@compiler-errors`` something generally useful but also required for rust-lang#142789

Rollup merge of #142822 - oli-obk:const-partial-eq, r=fee1-dead Make `PartialEq` a `const_trait` r? ``@fee1-dead`` or ``@compiler-errors`` something generally useful but also required for #142789

bors · 2025-06-23T16:57:20Z

☔ The latest upstream changes (presumably #142906) made this pull request unmergeable. Please resolve the merge conflicts.

compiler/rustc_codegen_cranelift/src/constant.rs

rustbot · 2025-06-27T13:58:35Z

The Miri subtree was changed

cc @rust-lang/miri

bors · 2025-06-27T20:19:12Z

☔ The latest upstream changes (presumably #143091) made this pull request unmergeable. Please resolve the merge conflicts.

RalfJung · 2025-06-28T12:41:39Z

@rustbot author

rustbot · 2025-06-28T12:41:43Z

Reminder, once the PR becomes ready for a review, use @rustbot ready.

oli-obk · 2025-06-30T15:34:53Z

Having provenance on some bytes still means we'd appropriately error out if someone just transmutes the entire thing to u128 in const-eval, or something like that. We'd only be truly exposing the "tail", i.e., everything after the first ptr-sized chunk.

I implemented this in 8a42250

It made things a lot cleaner and simpler

@rustbot ready

bors · 2025-07-03T08:15:48Z

☔ The latest upstream changes (presumably #143350) made this pull request unmergeable. Please resolve the merge conflicts.

RalfJung

Okay yes I think I can live with this approach. :)

RalfJung · 2025-07-03T19:54:40Z

compiler/rustc_const_eval/src/interpret/intrinsics.rs

+    let size = Size::from_bytes(16);
+    let align = tcx.data_layout.pointer_align;
+    let mut alloc = Allocation::new(size, *align, AllocInit::Uninit, ());


Any chance we can assert that this matches the size and align of the real TypeId?

RalfJung · 2025-07-03T19:59:15Z

compiler/rustc_const_eval/src/interpret/intrinsics.rs

@@ -29,6 +31,37 @@ pub(crate) fn alloc_type_name<'tcx>(tcx: TyCtxt<'tcx>, ty: Ty<'tcx>) -> ConstAll
    tcx.mk_const_alloc(alloc)
 }

+pub(crate) fn alloc_type_id<'tcx>(tcx: TyCtxt<'tcx>, ty: Ty<'tcx>) -> AllocId {


Suggested change

pub(crate) fn alloc_type_id<'tcx>(tcx: TyCtxt<'tcx>, ty: Ty<'tcx>) -> AllocId {

/// This returns the `AllocId` of a place where a [`TypeId`](https://doc.rust-lang.org/nightly/std/any/struct.TypeId.html) for the given `ty` is stored.

pub(crate) fn alloc_type_id<'tcx>(tcx: TyCtxt<'tcx>, ty: Ty<'tcx>) -> AllocId {

I think ideally this would return an OpTy rather than an AllocId, makes it conceptually much more clear -- we don't even want a place here.

Or, if we want to optimize for performance, this function should take as argument an MPlaceTy for where to store the TypeId.

I have a local commit for making it write to a dest directly. Which is a change I want to do for all the intrinsics that are currently generating a ConstValue and writing that from within the interpreter

I can land that first on master if you prefer. Or just move the commit for typeid into this PR

RalfJung · 2025-07-03T19:59:41Z

compiler/rustc_const_eval/src/interpret/intrinsics.rs

+
+    // Give the first pointer-size bytes provenance that knows about the type id
+
+    let alloc_id = tcx.reserve_and_set_type_id_alloc(ty);


There's two AllocId in this function, this one and the one being returned. Please give it a more clearly distinguished name.

RalfJung · 2025-07-03T20:01:02Z

compiler/rustc_const_eval/src/interpret/validity.rs

                                }
                            }
                        }
+                        // These are controlled by rustc and not available for CTFE
+                        GlobalAlloc::Type { .. } => skip_recursive_check = true,


Why is this need for Type but not for functions and vtables?

RalfJung · 2025-07-03T20:02:42Z

compiler/rustc_codegen_llvm/src/common.rs

+                    GlobalAlloc::Type { .. } => {
+                        // Drop the provenance, the offset contains the bytes of the hash
+                        let llval = self.const_usize(offset.bytes());
+                        return unsafe { llvm::LLVMConstIntToPtr(llval, llty) };


Why do we have to emit an int-to-ptr cast here? I think we should avoid that if at all possible...

RalfJung · 2025-07-03T20:10:21Z

src/tools/miri/src/intrinsics/mod.rs

@@ -461,6 +461,28 @@ pub trait EvalContextExt<'tcx>: crate::MiriInterpCxExt<'tcx> {
                // Make these a NOP, so we get the better Miri-native error messages.
            }

+            "type_id_eq" => {


I think what this should do is actually look up the two first pointers, get the corresponding Ty, and compare those. That also ensures that these are indeed (looking like) TypeId.

We can then also compare the raw bytes to catch programs doing nasty things, but comparing the actual true types is the main operation IMO. I'd rather not compare provenance as that's abstract data that cannot exist at runtime (as opposed to the Type allocations which we actually could manifest).

Please add a Miri test that transmutes two u128 to TypeId and compares them; that should be UB.

RalfJung · 2025-07-03T20:11:17Z

tests/mir-opt/gvn_type_id_polymorphic.rs

 const fn type_id_of_val<T: 'static>(_: &T) -> u128 {
-    std::intrinsics::type_id::<T>()
+    let name = std::intrinsics::type_name::<T>();
+    let len = name.len() as u64;
+    let len = u64::to_be_bytes(len);
+    let mut ret = [0; 16];
+    let mut i = 0;
+    while i < 8 {
+        ret[i] = len[i];
+        i += 1;
+    }
+    while i < 16 {
+        ret[i] = name.as_bytes()[i - 8];
+        i += 1;
+    }
+    u128::from_be_bytes(ret)
 }


What is this doing...? This definitely cannot stand without a comment.^^

RalfJung · 2025-07-03T20:11:53Z

tests/ui/consts/const_transmute_type_id.rs

+
+const _: () = {
+    let id = TypeId::of::<u8>();
+    let id: u8 = unsafe { (&id as *const TypeId).cast::<u8>().read() };


Suggested change

let id: u8 = unsafe { (&id as *const TypeId).cast::<u8>().read() };

let id: u8 = unsafe { (&raw const id).cast::<u8>().read() };

RalfJung · 2025-07-03T20:12:59Z

compiler/rustc_const_eval/src/const_eval/machine.rs

@@ -403,6 +403,22 @@ impl<'tcx> interpret::Machine<'tcx> for CompileTimeMachine<'tcx> {
                let cmp = ecx.guaranteed_cmp(a, b)?;
                ecx.write_scalar(Scalar::from_u8(cmp), dest)?;
            }
+            sym::type_id_eq => {


I think const-eval and Miri can use the same implementation.

RalfJung · 2025-07-03T20:13:44Z

library/core/src/any.rs

+
+            #[inline]
+            fn rt(a: &TypeId, b: &TypeId) -> bool {
+                a.data == b.data


This means Miri never calls the intrinsic, right?

rustbot assigned RalfJung Jun 20, 2025

This comment has been minimized.

Sign in to view

Dante-Broggi reviewed Jun 21, 2025

View reviewed changes

library/core/src/intrinsics/mod.rs Outdated Show resolved Hide resolved

Noratrieb reviewed Jun 21, 2025

View reviewed changes

library/core/src/any.rs Outdated Show resolved Hide resolved

oli-obk force-pushed the constable-type-id branch from 3cddd21 to 1fd7b66 Compare June 21, 2025 10:20

This comment has been minimized.

Sign in to view

oli-obk mentioned this pull request Jun 21, 2025

Make PartialEq a const_trait #142822

Merged

This comment has been minimized.

Sign in to view

antoyo reviewed Jun 21, 2025

View reviewed changes

compiler/rustc_codegen_gcc/src/common.rs Outdated Show resolved Hide resolved

bjorn3 reviewed Jun 21, 2025

View reviewed changes

compiler/rustc_codegen_cranelift/src/constant.rs Outdated Show resolved Hide resolved

oli-obk force-pushed the constable-type-id branch 2 times, most recently from b8a7a10 to 1c47a64 Compare June 24, 2025 09:25

bjorn3 reviewed Jun 24, 2025

View reviewed changes

compiler/rustc_codegen_cranelift/src/constant.rs Outdated Show resolved Hide resolved

This comment has been minimized.

Sign in to view

oli-obk force-pushed the constable-type-id branch from 1c47a64 to bcb4aa2 Compare June 24, 2025 13:02

This comment has been minimized.

Sign in to view

bjorn3 reviewed Jun 24, 2025

View reviewed changes

compiler/rustc_codegen_cranelift/src/constant.rs Outdated Show resolved Hide resolved

oli-obk force-pushed the constable-type-id branch from afa74de to 6fea6c3 Compare June 27, 2025 13:58

This comment has been minimized.

Sign in to view

oli-obk force-pushed the constable-type-id branch from 6fea6c3 to 2e8cece Compare June 27, 2025 14:24

This comment has been minimized.

Sign in to view

rustbot removed the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Jun 28, 2025

rustbot added the S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. label Jun 28, 2025

oli-obk force-pushed the constable-type-id branch from 2e8cece to aa4c156 Compare June 30, 2025 13:19

This comment has been minimized.

Sign in to view

oli-obk force-pushed the constable-type-id branch from aa4c156 to 8a42250 Compare June 30, 2025 15:34

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Jun 30, 2025

This comment has been minimized.

Sign in to view

fasterthanlime mentioned this pull request Jul 1, 2025

Introduce ShapeId facet-rs/facet#540

Closed

oli-obk force-pushed the constable-type-id branch from 8a42250 to 82aa2a9 Compare July 1, 2025 14:35

oli-obk added 8 commits July 3, 2025 08:22

Add opaque TypeId handles for CTFE

f7ab598

Change type id to contain a pointer to the unique type id information

52b9ece

Some simplifications

b16524d

Make TypeId robust against compile-time introspection

552ba12

Make TypeId const-comparable

bc9d010

Split up TypeId into pointer sized fields for CTFE to use as opaque ids

8e4dec6

Cursed gvn type_id test

7404aa3

Only mark the first pointer-width bytes of TypeIdwith provenance

871e4eb

oli-obk force-pushed the constable-type-id branch from 82aa2a9 to 871e4eb Compare July 3, 2025 08:22

RalfJung reviewed Jul 3, 2025

View reviewed changes

	pub(crate) fn alloc_type_id<'tcx>(tcx: TyCtxt<'tcx>, ty: Ty<'tcx>) -> AllocId {
	/// This returns the `AllocId` of a place where a [`TypeId`](https://doc.rust-lang.org/nightly/std/any/struct.TypeId.html) for the given `ty` is stored.
	pub(crate) fn alloc_type_id<'tcx>(tcx: TyCtxt<'tcx>, ty: Ty<'tcx>) -> AllocId {


		// Give the first pointer-size bytes provenance that knows about the type id

		let alloc_id = tcx.reserve_and_set_type_id_alloc(ty);

	let id: u8 = unsafe { (&id as *const TypeId).cast::<u8>().read() };
	let id: u8 = unsafe { (&raw const id).cast::<u8>().read() };

Make TypeId const comparable #142789

Are you sure you want to change the base?

Make TypeId const comparable #142789

Conversation

oli-obk commented Jun 20, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rustbot commented Jun 20, 2025

Uh oh!

This comment has been minimized.

Uh oh!

Uh oh!

This comment has been minimized.

RalfJung commented Jun 21, 2025

Uh oh!

rustbot commented Jun 21, 2025

Uh oh!

oli-obk commented Jun 21, 2025

Uh oh!

oli-obk commented Jun 21, 2025

Uh oh!

This comment has been minimized.

Uh oh!

Uh oh!

bors commented Jun 23, 2025

Uh oh!

Uh oh!

This comment has been minimized.

This comment has been minimized.

Uh oh!

rustbot commented Jun 27, 2025

Uh oh!

This comment has been minimized.

This comment has been minimized.

bors commented Jun 27, 2025

Uh oh!

RalfJung commented Jun 28, 2025

Uh oh!

rustbot commented Jun 28, 2025

Uh oh!

This comment has been minimized.

oli-obk commented Jun 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment has been minimized.

bors commented Jul 3, 2025

Uh oh!

RalfJung left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

oli-obk commented Jun 20, 2025 •

edited

Loading

oli-obk commented Jun 30, 2025 •

edited

Loading