const-eval error: always say in which item the error occurred #142008

RalfJung · 2025-06-04T07:14:22Z

I don't see why "is this generic" should make a difference. It may be reasonable to key this on whether the error occurs in a const fn that was invoked by a const (making it non-obvious which constant it is) vs inside the body of the const.

r? @oli-obk

rustbot · 2025-06-04T07:14:28Z

Some changes occurred to the CTFE machinery

cc @RalfJung, @oli-obk, @lcnr

RalfJung · 2025-06-04T07:15:50Z

tests/ui/const-generics/defaults/default-param-wf-concrete.next.stderr

@@ -2,7 +2,7 @@ error[E0080]: attempt to compute `u8::MAX + 1_u8`, which would overflow
  --> $DIR/default-param-wf-concrete.rs:4:28
   |
 LL | struct Foo<const N: u8 = { 255 + 1 }>;
-   |                            ^^^^^^^ evaluation of constant value failed here
+   |                            ^^^^^^^ evaluation of `Foo::{constant#0}` failed here


This isn't great, ideally it'd say something like "evaluation of unnamed constant inside Foo failed here" or so...

yea, this may be sth we want to improve for all mentions of unnamed consts. The disambiguators don't help anyone but compiler devs.

RalfJung · 2025-06-04T07:16:13Z

tests/ui/borrowck/issue-81899.stderr

@@ -2,7 +2,7 @@ error[E0080]: evaluation panicked: explicit panic
  --> $DIR/issue-81899.rs:6:24
   |
 LL | const _CONST: &[u8] = &f(&[], |_| {});
-   |                        ^^^^^^^^^^^^^^ evaluation of constant value failed here
+   |                        ^^^^^^^^^^^^^^ evaluation of `f::<{closure@$DIR/issue-81899.rs:6:31: 6:34}>` failed here


I am not entirely sure why this doesn't say _CONST...

Uh, we take the instance from the top of the stack. Possibly a remnant from early miri days? We should just take the current item's global id

I haven't found where in the code we do that...

Nope, we don't. huh. This is very weird. Even if some promotion thing were going on, we should be reporting the promoted in that case

Right, you're sharing my confusion now. ;)

I can only look at this more closely next weekend -- I thought this would be a largely editorial change.^^

lmao I found it

rust/compiler/rustc_const_eval/src/errors.rs

Lines 286 to 290 in b331b8b

impl Subdiagnostic for FrameNote {

fn add_to_diag<G: EmissionGuarantee>(self, diag: &mut Diag<'_, G>) {

diag.arg("times", self.times);

diag.arg("where_", self.where_);

diag.arg("instance", self.instance);

Oh wow, this is nasty form of global state...

RalfJung · 2025-06-04T07:17:56Z

tests/ui/const-generics/issues/issue-100313.stderr

@@ -2,7 +2,7 @@ error[E0080]: writing to ALLOC0 which is read-only
  --> $DIR/issue-100313.rs:18:5
   |
 LL |     x.set_false();
-   |     ^^^^^^^^^^^^^ evaluation of constant value failed here
+   |     ^^^^^^^^^^^^^ evaluation of `T::<&true>::set_false` failed here


Why does this use the name of the function rather than the name of the surrounding constant...?

RalfJung · 2025-06-04T07:21:04Z

tests/ui/consts/const-eval/const_fn_ptr_fail2.stderr

@@ -14,7 +14,7 @@ error[E0080]: calling non-const function `double`
  --> $DIR/const_fn_ptr_fail2.rs:19:18
   |
 LL | const Z: usize = bar(double, 2); // FIXME: should fail to typeck someday
-   |                  ^^^^^^^^^^^^^^ evaluation of constant value failed here
+   |                  ^^^^^^^^^^^^^^ evaluation of `bar` failed here


Okay I think this is just using the wrong instance? It seems to be the instance of the top frame, rather than the surrounding constant.

I entirely forgot that we print the backtrace backwards now. Still not sure if that's really a good idea. But that means the error itself always points inside the constant, right? So maybe there's no reason to put the constant name in the error message after all.

Yeah something is very odd here: #142010

rust-log-analyzer · 2025-06-04T07:51:10Z

The job x86_64-gnu-tools failed! Check out the build log: (web) (plain)

Click to see the possible cause of the failure (guessed by this bot)

tests/fail/function_calls/exported_symbol_bad_unwind2.rs (revision `definition`) ... ok
tests/fail/function_calls/exported_symbol_bad_unwind2.rs (revision `both`) ... ok

FAILED TEST: tests/fail/const-ub-checks.rs
command: MIRI_ENV_VAR_TEST="0" MIRI_TEMP="/tmp/miri-uitest-9em7Vp" RUST_BACKTRACE="1" "/checkout/obj/build/x86_64-unknown-linux-gnu/stage2/bin/miri" "--error-format=json" "--sysroot=/checkout/obj/build/x86_64-unknown-linux-gnu/miri-sysroot" "-Dwarnings" "-Dunused" "-Ainternal_features" "-Zui-testing" "--out-dir" "/checkout/obj/build/x86_64-unknown-linux-gnu/stage1-tools/x86_64-unknown-linux-gnu/tmp/miri_ui/0/tests/fail" "tests/fail/const-ub-checks.rs" "--edition" "2021"

error: actual output differed from expected
Execute `./miri test --bless` to update `tests/fail/const-ub-checks.stderr` to the actual output
--- tests/fail/const-ub-checks.stderr
+++ <stderr output>
---
   |
LL |     let _x = UNALIGNED_READ;
   |              ^^^^^^^^^^^^^^
   |
   = note: duplicate diagnostic emitted due to `-Z deduplicate-diagnostics=no`

error: aborting due to 1 previous error

For more information about this error, try `rustc --explain E0080`.

---
   |
LL |     let _x = UNALIGNED_READ;
   |              ^^^^^^^^^^^^^^
   |
   = note: duplicate diagnostic emitted due to `-Z deduplicate-diagnostics=no`

error: aborting due to 1 previous error

For more information about this error, try `rustc --explain E0080`.

full stdout:



FAILED TEST: tests/fail/erroneous_const2.rs
command: MIRI_ENV_VAR_TEST="0" MIRI_TEMP="/tmp/miri-uitest-9em7Vp" RUST_BACKTRACE="1" "/checkout/obj/build/x86_64-unknown-linux-gnu/stage2/bin/miri" "--error-format=json" "--sysroot=/checkout/obj/build/x86_64-unknown-linux-gnu/miri-sysroot" "-Dwarnings" "-Dunused" "-Ainternal_features" "-Zui-testing" "--out-dir" "/checkout/obj/build/x86_64-unknown-linux-gnu/stage1-tools/x86_64-unknown-linux-gnu/tmp/miri_ui/0/tests/fail" "tests/fail/erroneous_const2.rs" "--edition" "2021"

error: actual output differed from expected
Execute `./miri test --bless` to update `tests/fail/erroneous_const2.stderr` to the actual output
--- tests/fail/erroneous_const2.stderr
+++ <stderr output>
---
   |
LL |     println!("{}", FOO);
   |                    ^^^
   |
   = note: duplicate diagnostic emitted due to `-Z deduplicate-diagnostics=no`

note: erroneous constant encountered
  --> tests/fail/erroneous_const2.rs:8:20
   |
LL |     println!("{}", FOO);
---
   |
LL |     println!("{}", FOO);
   |                    ^^^
   |
   = note: duplicate diagnostic emitted due to `-Z deduplicate-diagnostics=no`

note: erroneous constant encountered
  --> tests/fail/erroneous_const2.rs:8:20
   |
LL |     println!("{}", FOO);
---

Location:
   /cargo/registry/src/index.crates.io-1949cf8c6b5b557f/ui_test-0.29.2/src/lib.rs:369

Backtrace omitted. Run with RUST_BACKTRACE=1 environment variable to display it.
Run with RUST_BACKTRACE=full to include source snippets.
error: test failed, to rerun pass `--test ui`

Caused by:
  process didn't exit successfully: `/checkout/obj/build/x86_64-unknown-linux-gnu/stage1-tools/x86_64-unknown-linux-gnu/release/deps/ui-f58affc77002becb` (exit status: 1)
Command has failed. Rerun with -v to see more details.
Build completed unsuccessfully in 0:05:06
  local time: Wed Jun  4 07:50:36 UTC 2025
  network time: Wed, 04 Jun 2025 07:50:36 GMT
##[error]Process completed with exit code 1.
Post job cleanup.

RalfJung · 2025-06-04T09:14:11Z

Once #142015 lands, do we even still want this? The first label always points inside the "root constant" (I forgot about that), so we don't have to repeat the name of that constant.

It's kind of unnecessary that we spell out whether it's a static or const but 🤷

oli-obk · 2025-06-04T09:49:34Z

I think we still want this PR just as a simplification without any downsides

bors · 2025-06-05T20:47:46Z

☔ The latest upstream changes (presumably #142081) made this pull request unmergeable. Please resolve the merge conflicts.

const-eval error: always say in which item the error occurred

6a9919e

rustbot assigned oli-obk Jun 4, 2025

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. labels Jun 4, 2025

RalfJung commented Jun 4, 2025

View reviewed changes

	impl Subdiagnostic for FrameNote {
	fn add_to_diag<G: EmissionGuarantee>(self, diag: &mut Diag<'_, G>) {
	diag.arg("times", self.times);
	diag.arg("where_", self.where_);
	diag.arg("instance", self.instance);

const-eval error: always say in which item the error occurred #142008

Are you sure you want to change the base?

const-eval error: always say in which item the error occurred #142008

Conversation

RalfJung commented Jun 4, 2025

Uh oh!

rustbot commented Jun 4, 2025

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!

Choose a reason for hiding this comment

Uh oh!

rust-log-analyzer commented Jun 4, 2025

Uh oh!

RalfJung commented Jun 4, 2025

Uh oh!

oli-obk commented Jun 4, 2025

Uh oh!

bors commented Jun 5, 2025

Uh oh!

Uh oh!