Roll toolchain to nightly-2022-11-03 by randomPoison · Pull Request #1614 · immunant/c2rust

randomPoison · 2026-02-25T00:39:19Z

Copilot choked and died on this, but Codex seems to have succeeded. I let it spin for like an hour and it got code compiling and tests running, so let's let CI run and see if anything is broken that I didn't catch locally.

Roll the nightly toolchain 6 weeks forward to nightly-2022-11-03. I've reviewed and cleaned up the changes and they all seem reasonable to me now, so this is ready for review.

Rua · 2026-03-03T11:45:48Z

Why was that particular nightly version chosen and not, say, a much newer one?

kkysen

Why was that particular nightly version chosen and not, say, a much newer one?

We were mostly just testing out codex on this. We decided to first start with a smaller jump so that it could handle it better. We may try a longer jump next, but there's a larger risk things go wrong, and in case we introduce some regression, it'll be harder to debug/rollback, and a much larger set of changes will be much more difficult to review as well.

And also add the new Assume intrinsic while we're at it.

kkysen · 2026-03-17T05:52:22Z

-            ExprKind::MethodCall(_seg, args, _span) => {
+            ExprKind::MethodCall(_seg, recv, args, _span) => {
                if let Some(fn_sig) = opt_fn_sig {
+                    if let Some(&ty) = fn_sig.inputs().get(0) {
+                        illtyped |= self.ensure(recv, ty);
+                    }
                    for (i, arg) in args.iter_mut().enumerate() {
-                        if let Some(&ty) = fn_sig.inputs().get(i) {
+                        if let Some(&ty) = fn_sig.inputs().get(i + 1) {
                            illtyped |= self.ensure(arg, ty);


We have a bunch of different ways of handling the recv/args split. Could we try to collapse that to just one? That is, have a fn combine_recv_and_args(recv: P<Expr>, args: Vec<P<Expr>>) -> impl Iterator<Item = P<Expr>> that we use everywhere. We can iterate over it, use .zip (e.g., here with fn_sig.inputs()), .nth (would work here, too), .enumerate, which I think should cover almost all of the use cases and keeps the usage much more standardized.

Hm, I don't think we need a helper function here, I think the common pattern of iter::once(recv).chain(args) is good enough. I did a pass over the places where we're handling MethodCall and unified a couple to use that approach.

ahomescu · 2026-03-18T04:12:02Z

                            }
                            Rvalue::Cast(
-                                CastKind::PtrToPtr | CastKind::FnPtrToPtr,
+                                CastKind::PtrToPtr | CastKind::FnPtrToPtr | CastKind::IntToInt,


Is this correct? IntToInt might truncate or extend.

The comment below this specifically mentions covering casts like usize to size_t, so this seems reasonable, but without spending more time digging I can't be sure.

ahomescu · 2026-04-20T23:40:58Z

@@ -260,7 +260,7 @@ fn run_polonius<'tcx>(
    //pretty::write_mir_fn(tcx, mir, &mut |_, _| Ok(()), &mut std::io::stdout()).unwrap();


It would save us a lot of review time (and the agent itself a lot of cycles) if it updated every c2rust component separately next time. Then we could prioritize c2rust-refactor higher, and leave the others for later (I don't know how much we care about the ALLSTAR stuff right now).

Yeah I agree. I think before attempting larger toolchain rolls we need to trim down the code and disable stuff we don't actively want to support. There are a lot of changes in code that we're effectively treating as dead, and that adds a lot of noise to the PR. We also need to revive more of the c2rust-refactor test suite, since there are a bunch of transforms that are currently dead, making changes to that code hard to verify and thus harder to review.

ahomescu · 2026-04-21T00:37:15Z

        // We do not have a SpanNodeKind for certain nodes
        Some(Node::TypeBinding(_)) => None,
        Some(Node::TraitRef(_)) => None,
+        Some(Node::ExprField(_)) => None,


These aren't technically FieldDefs. ExprField and PatField contain an Expr and Pat, respectively, so I'm wondering if we do need a new NodeSpanKind here or two. Maybe one shared between the two of them:

Some(Node::ExprField(expr)) => Some(NodeSpan::new(expr.span, Field)), Some(Node::PatField(pat)) => Some(NodeSpan::new(pat.span, Field)),

since ExprField and PatField can't collide with each other, but they may collide with their inner values.

Otoh since ExprField is going to be something like ident: value, maybe there isn't the chance of a collision. Do we have a test for these?

Hm, I don't think we need new node types, we already have ExprField and NodeField that we were using in a couple of places. I've updated this function to return span info for those.

What exactly do you mean by "collision" here? Are you referring to overlapping spans? My understanding is that ExprField and PatField can't overlap with each other because they show up in different positions (expressions vs patterns), but maybe you're referring to something else.

As far as I know none of our existing transforms hit this code path, so we don't have any tests to cover it. I can try adding one, but I'm hesitant to put together a contrived test for this functionality. I'd rather have tests for real transforms that naturally hit this code path, and I don't see a good option for that currently.

Are you referring to overlapping spans?

That's right, we were getting multiple nodes with identical spans because sometimes an AST node will include an inner node with the same span. For example, a PathExpr and the inner Path node overlap.

My understanding is that ExprField and PatField can't overlap with each other

That's right, but I was worried that ExprField and the inner Expr might collide. On second thought, that's probably not the case.

Would field shorthand e.g. in let field = ...; FooStruct { field } cause the spans to collide? The shorthand is usable in both expressions and patterns.

kkysen

There's one more nightly-2022-08-08 left in c2rust-rust-tools/src/lib.rs. Otherwise, I think it LGTM now.

randomPoison · 2026-04-22T23:50:14Z

@kkysen Ah right good catch. I updated the comment to not specifically reference the exact nightly, I'd prefer that we don't have to keep updating references to the exact nightly every time we do a toolchain roll.

kkysen · 2026-04-22T23:53:50Z

I'd prefer that we don't have to keep updating references to the exact nightly every time we do a toolchain roll.

I usually do because it makes it very easy to find. For example, once that nightly toolchain reaches edition 2024, that comment should be removed/changed, but it's now harder to find it.

randomPoison · 2026-04-23T00:00:12Z

Ah sure, I see how that'd be useful. Changed it to keep the comment and just update the nightly name.

randomPoison added 7 commits February 24, 2026 14:42

First pass of cleanup bumping to nightly-2022-11-03

ca37d6a

Update ast.txt and friends

17d9435

Fix remaining compile errors

6f0e762

Minor fix in ast.txt

b295d6b

Fix c2rust-analyze tests

688b276

Run cargo fmt

4283bbf

Fix warnings

a50e57d

ahomescu reviewed Feb 25, 2026

View reviewed changes

kkysen reviewed Mar 3, 2026

View reviewed changes

Rua mentioned this pull request Mar 9, 2026

Clarification of stance on AI use #1653

Open

randomPoison added 7 commits March 10, 2026 10:17

Add TODO comment

a0436f5

Merge branch 'master' into legare/codex-bump-nightly-experiment

1855356

Reapply toolchain updates after merge

3e94d9e

Update type cast check

69296d9

Cleanup method call args

f417903

Ignore c2rust-analyze test binaries

5c3f68d

Add back handling of CopyNonOverlapping intrinsic

017cb29

And also add the new Assume intrinsic while we're at it.

randomPoison force-pushed the legare/codex-bump-nightly-experiment branch from 6efc208 to 017cb29 Compare March 10, 2026 22:16

randomPoison added 8 commits March 10, 2026 15:19

Run cargo fmt

52838ed

Add todo! for OpaqueCast

cb7cb79

Fix the c2rust-analyze tests in a more minimal way

716dd53

Leave Intrinsic statements unimplemented for now

b7e1e7b

Fix mutability handling in ident_pat

b37957c

Add snapshot test for literal matchers

034a8d9

Suppport substituting matched literals

a218fa7

Add test for matching a mut pattern

6ab1a49

randomPoison commented Mar 12, 2026

View reviewed changes

Comment thread c2rust-refactor/src/ast_manip/util.rs

randomPoison added 2 commits March 12, 2026 15:29

Use shorthand syntax because it's more idiomatic

4c2cfdc

Don't duplicate the logic from handle_seq

10a8005

kkysen reviewed Mar 17, 2026

View reviewed changes

Comment thread c2rust-refactor/src/collapse/deleted.rs Outdated

kkysen reviewed Mar 17, 2026

View reviewed changes

kkysen requested changes Mar 17, 2026

View reviewed changes

kkysen reviewed Mar 17, 2026

View reviewed changes

Comment thread c2rust-refactor/src/transform/literals.rs Outdated

Don't support dyn casts

fe20ad4

ahomescu reviewed Mar 18, 2026

View reviewed changes

randomPoison added 10 commits April 20, 2026 11:56

c2rust-analyze: Fix operands for CopyNonOverlapping

cc85c89

c2rust-refactor: Remove unnecessary .iter() call

b0cca55

Cleanup MethodCall handling in deleted.rs

fc6d821

Normalize handling of MethodCall recv and args a bit

1fcbbed

Use size hint in handle_seq

c425aa8

Normalize name of MethodCall recv + args

1a175c5

Update remaining references to old toolchain

a1e085b

Fix rewrite_stmts snapshot test

d464dac

Merge branch 'master' into legare/codex-bump-nightly-experiment

d745787

Fix a feature that somehow destabilized during the toolchain roll

bc0ee23

ahomescu reviewed Apr 21, 2026

View reviewed changes

randomPoison added 3 commits April 21, 2026 11:31

Update references to the current rust version in a few places

06fb3c8

Add todo! instead of TODO comment

88ad88c

Generate node spans for ExperField and PatField

e6c6ebc

ahomescu approved these changes Apr 21, 2026

View reviewed changes

kkysen requested changes Apr 22, 2026

View reviewed changes

Fix the last reference to the old nightly

06dcec8

randomPoison force-pushed the legare/codex-bump-nightly-experiment branch from 166550a to 06dcec8 Compare April 23, 2026 00:00

kkysen approved these changes Apr 23, 2026

View reviewed changes

randomPoison merged commit 0389785 into master Apr 23, 2026
18 of 20 checks passed

randomPoison deleted the legare/codex-bump-nightly-experiment branch April 23, 2026 19:42

		@@ -260,7 +260,7 @@ fn run_polonius<'tcx>(
		//pretty::write_mir_fn(tcx, mir, &mut \|_, _\| Ok(()), &mut std::io::stdout()).unwrap();

Conversation

randomPoison commented Feb 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Rua commented Mar 3, 2026

Uh oh!

kkysen left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

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!

kkysen left a comment

Choose a reason for hiding this comment

Uh oh!

randomPoison commented Apr 22, 2026

Uh oh!

kkysen commented Apr 22, 2026

Uh oh!

randomPoison commented Apr 23, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

randomPoison commented Feb 25, 2026 •

edited

Loading