Postpone the evaluation of constant expressions that depend on inference variables by b-naber · Pull Request #90023 · rust-lang/rust

b-naber · 2021-10-18T16:40:04Z

Previously delay_span_bug calls were triggered once an inference variable was included in the substs of a constant that was to be evaluated. Some of these would merely have resulted in trait candidates being rejected, hence no real error was ever encountered, but the triggering of the delay_span_bug then caused an ICE in later stages of the compiler due to no error ever occurring.
We now postpone the evaluation of these constants, so any trait obligation fulfillment will simply stall on this constant and the existing type inference machinery of the compiler handles any type errors if present.

Fixes#89320
Fixes#89146
Fixes#87964
Fixes#87470
Fixes#83288
Fixes#83249
Fixes#90654

I want to thank @BoxyUwU for cooperating on this and for providing some help.

r? @lcnr maybe?

rust-highfive · 2021-10-18T16:40:07Z

Some changes occured to the CTFE / Miri engine

cc @rust-lang/miri

BoxyUwU · 2021-10-18T19:03:00Z

I think we should be erasing lifetimes in param_env and substs at the start of this fn

y, with this PR we can erase lifetimes and don't have to canonicalize anymore

Ok, I'll change that. Can you explain why we should do this? Is it just that we don't need any lifetime information anymore during codegen (which the evaluation of constants loosely speaking is)?

Is it just that we don't need any lifetime information anymore during codegen
yes, const eval and codegen (which should pretty much behave the same) do not depend on lifetime information so it is fine to erase them. That also improves cashing as all lifetime params are always 'erased.

lcnr · 2021-10-28T04:41:51Z

forgot to write this when I first reviewed this:

the change looks good to me but I think we should not land this if we ever intend to revert this PR, as in that case it simply masks bugs we will have to fix then. My only concern here is that anonymous constants can refer to unused generic arguments, so by not trying to evaluate them we will probably prevent some really rare cases from compiling. Need to take some time to figure out the greater picture here, so it might take a bit until this PR gets merged.

nikomatsakis

r=me with the change below

nikomatsakis · 2021-11-04T18:23:16Z

I suggest we re-order this as follows:
Suggested change
let param_env = self.tcx.erase_regions(param_env);
letmut substs = unevaluated.substs(self.tcx);
substs = self.tcx.erase_regions(substs);
substs = self.resolve_vars_if_possible(substs);
// Postpone the evaluation of constants whose substs depend on inference
// variables
if substs.has_infer_types_or_consts(){
returnErr(ErrorHandled::TooGeneric);
}
let unevaluated = ty::Unevaluated{
def: unevaluated.def,
substs_:Some(substs),
promoted: unevaluated.promoted,
};
letmut substs = unevaluated.substs(self.tcx);
substs = self.resolve_vars_if_possible(substs);
// Postpone the evaluation of constants whose substs depend on inference
// variables
if substs.has_infer_types_or_consts(){
returnErr(ErrorHandled::TooGeneric);
}
let param_env_erased = self.tcx.erase_regions(param_env);
let substs_erased = self.tcx.erase_regions(substs);
let unevaluated = ty::Unevaluated{
def: unevaluated.def,
substs_:Some(substs_erased),
promoted: unevaluated.promoted,
};
// The return value is the evaluated value which doesn't contain any reference to inference
// variables, thus we don't need to substitute back the original values.
self.tcx.const_eval_resolve(param_env, unevaluated, span)
It is necessary to resolve variables before erasing because an inference variable may resolve to an &T of some kind.

…s or regions

jackh726 · 2021-12-04T21:09:11Z

@bors r=nikomatsakis

bors · 2021-12-04T21:09:13Z

📌 Commit 37ed2db has been approved by nikomatsakis

…askrgr Rollup of 6 pull requests Successful merges: - rust-lang#89642 (environ on macos uses directly libc which has the correct signature.) - rust-lang#90022 (Explain why `Self` is invalid in generic parameters) - rust-lang#90023 (Postpone the evaluation of constant expressions that depend on inference variables) - rust-lang#91215 (Implement VecDeque::retain_mut) - rust-lang#91355 (std: Stabilize the `thread_local_const_init` feature) - rust-lang#91528 (LLVM support .insn directive) Failed merges: r? `@ghost` `@rustbot` modify labels: rollup

Alexendoo · 2021-12-06T13:23:33Z

The following also look to be fixed by this PR 🎉

I haven't closed or marked any as needs-test, since I'm not sure which would be considered duplicates

jackh726 · 2021-12-07T22:27:40Z

@b-naber any chance I can enlist you to go through the above issues and decide which are "duplicates" and which need a test?

b-naber · 2021-12-07T22:36:56Z

@b-naber any chance I can enlist you to go through the above issues and decide which are "duplicates" and which need a test?

Of course, I'll do that tomorrow.

…al, r=jackh726 Add tests fixed by rust-lang#90023 The following issues were fixed by rust-lang#90023Fixesrust-lang#79674Fixesrust-lang#83765Fixesrust-lang#86033Fixesrust-lang#90318Fixesrust-lang#88468 The following issues were duplicates of rust-lang#90654Fixesrust-lang#86850Fixesrust-lang#89022 r? `@jackh726`

…al, r=jackh726 Add tests fixed by rust-lang#90023 The following issues were fixed by rust-lang#90023Fixesrust-lang#79674Fixesrust-lang#83765Fixesrust-lang#86033Fixesrust-lang#90318Fixesrust-lang#88468 The following issues were duplicates of rust-lang#90654Fixesrust-lang#86850Fixesrust-lang#89022 r? ``@jackh726``

…al, r=jackh726 Add tests fixed by rust-lang#90023 The following issues were fixed by rust-lang#90023Fixesrust-lang#79674Fixesrust-lang#83765Fixesrust-lang#86033Fixesrust-lang#90318Fixesrust-lang#88468 The following issues were duplicates of rust-lang#90654Fixesrust-lang#86850Fixesrust-lang#89022 r? ```@jackh726```

…al, r=jackh726 Add tests fixed by rust-lang#90023 The following issues were fixed by rust-lang#90023Fixesrust-lang#79674Fixesrust-lang#83765Fixesrust-lang#86033Fixesrust-lang#90318Fixesrust-lang#88468 The following issues were duplicates of rust-lang#90654Fixesrust-lang#86850Fixesrust-lang#89022 r? ````@jackh726````

…askrgr Rollup of 11 pull requests Successful merges: - rust-lang#91668 (Remove the match on `ErrorKind::Other`) - rust-lang#91678 (Add tests fixed by rust-lang#90023) - rust-lang#91679 (Move core/stream/stream/mod.rs to core/stream/stream.rs) - rust-lang#91681 (fix typo in `intrinsics::raw_eq` docs) - rust-lang#91686 (Fix `Vec::reserve_exact` documentation) - rust-lang#91697 (Delete Utf8Lossy::from_str) - rust-lang#91706 (Add unstable book entries for parts of asm that are not being stabilized) - rust-lang#91709 (Replace iterator-based set construction by *Set::From<[T; N]>) - rust-lang#91716 (Improve x.py logging and defaults a bit more) - rust-lang#91747 (Add pierwill to .mailmap) - rust-lang#91755 (Fix since attribute for const_linked_list_new feature) Failed merges: r? `@ghost` `@rustbot` modify labels: rollup

…ce, r=lcnr Directly use ConstValue for single literals in blocks Addresses the minimal repro in rust-lang#92186, but doesn't fix the underlying problem (which would be solved by solving the anon subst problem afaict). I do, however, think that it makes sense in general to treat single literals in anon blocks as const values directly, especially in light of the problem that the issue refers to (anon const evaluation being postponed until infer variables in substs can be resolved, which was introduced by rust-lang#90023), i.e. while we do get warnings for those unnecessary braces, we should try to avoid errors caused by those braces if possible.

rust-highfive assigned lcnr Oct 18, 2021

rust-highfive added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Oct 18, 2021

This comment has been minimized.
Sign in to view

RalfJung reviewed Oct 18, 2021
View reviewed changes

Comment threadcompiler/rustc_infer/src/infer/mod.rs Outdated

BoxyUwU self-assigned this Oct 18, 2021

b-naber force-pushed the postpone_const_eval_infer_vars branch 2 times, most recently from effd201 to afa34ceCompare October 18, 2021 17:57

BoxyUwU requested changes Oct 18, 2021
View reviewed changes

oli-obk reviewed Oct 18, 2021
View reviewed changes

Comment threadsrc/test/ui/const-generics/generic_const_exprs/const_eval_resolve_canonical.rs Outdated

lcnr reviewed Oct 18, 2021
View reviewed changes

Comment threadsrc/test/ui/const-generics/issues/issue-89320.rs Outdated

b-naber force-pushed the postpone_const_eval_infer_vars branch from afa34ce to 42695c2Compare October 18, 2021 20:58

apiraino added the T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. label Oct 28, 2021

nikomatsakis reviewed Nov 4, 2021
View reviewed changes

This comment has been minimized.
Sign in to view

b-naber added 3 commits November 5, 2021 18:40

postpone evaluation of constants whose substs depend on inference var…
7530c43
…s or regions

add tests
6000b48

resolve variables before erasing lifetimes
dae2407

b-naber force-pushed the postpone_const_eval_infer_vars branch from effbfa8 to 3a7af1bCompare November 7, 2021 21:40

This comment has been minimized.
Sign in to view

consider unevaluated consts in extract_inference_diagnostics_data
37ed2db

b-naber force-pushed the postpone_const_eval_infer_vars branch from 3a7af1b to 37ed2dbCompare November 7, 2021 22:00

This was referenced Nov 8, 2021
ICE with basic generic const expr #90654
Closed
ICE with const generics: asked to assemble builtin bounds of unexpected type #79674
Closed

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Dec 4, 2021

matthiaskrgr mentioned this pull request Dec 4, 2021
Rollup of 6 pull requests #91539
Merged

bors merged commit 1f2a26e into rust-lang:masterDec 5, 2021

rustbot added this to the 1.59.0 milestone Dec 5, 2021

b-naber deleted the postpone_const_eval_infer_vars branch December 5, 2021 22:49

b-naber mentioned this pull request Dec 8, 2021
Add tests fixed by #90023 #91678
Merged

matthiaskrgr mentioned this pull request Dec 9, 2021
Rollup of 6 pull requests #91712
Closed

matthiaskrgr mentioned this pull request Dec 10, 2021
Rollup of 12 pull requests #91726
Closed

matthiaskrgr mentioned this pull request Dec 10, 2021
Rollup of 22 pull requests #91756
Closed

matthiaskrgr mentioned this pull request Dec 10, 2021
Rollup of 11 pull requests #91761
Merged

lcnr mentioned this pull request Dec 11, 2021
Tracking issue for const generics (RFC 2000) #44580
Closed
16 tasks

lcnr added the A-const-generics Area: const generics (parameters and arguments) label Dec 11, 2021

BoxyUwU mentioned this pull request Dec 22, 2021
Conflicting implementation of trait with generic_const_exprs feature #92186
Open

b-naber mentioned this pull request Jan 11, 2022
Directly use ConstValue for single literals in blocks #92780
Merged

-let param_env = self.tcx.erase_regions(param_env);
-letmut substs = unevaluated.substs(self.tcx);
- substs = self.tcx.erase_regions(substs);
- substs = self.resolve_vars_if_possible(substs);
-// Postpone the evaluation of constants whose substs depend on inference
-// variables
-if substs.has_infer_types_or_consts(){
-returnErr(ErrorHandled::TooGeneric);
-}
-let unevaluated = ty::Unevaluated{
- def: unevaluated.def,
-substs_:Some(substs),
- promoted: unevaluated.promoted,
-};
+letmut substs = unevaluated.substs(self.tcx);
+ substs = self.resolve_vars_if_possible(substs);
+// Postpone the evaluation of constants whose substs depend on inference
+// variables
+if substs.has_infer_types_or_consts(){
+returnErr(ErrorHandled::TooGeneric);
+}
+let param_env_erased = self.tcx.erase_regions(param_env);
+let substs_erased = self.tcx.erase_regions(substs);
+let unevaluated = ty::Unevaluated{
+ def: unevaluated.def,
+substs_:Some(substs_erased),
+ promoted: unevaluated.promoted,
+};
+// The return value is the evaluated value which doesn't contain any reference to inference
+// variables, thus we don't need to substitute back the original values.
+self.tcx.const_eval_resolve(param_env, unevaluated, span)

Uh oh!

Uh oh!

Conversation

b-naber commented Oct 18, 2021• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rust-highfive commented Oct 18, 2021

Uh oh!

This comment has been minimized.

Uh oh!

BoxyUwUOct 18, 2021

Choose a reason for hiding this comment

Uh oh!

lcnrOct 18, 2021• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

b-naberOct 18, 2021

Choose a reason for hiding this comment

Uh oh!

lcnrOct 18, 2021

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

lcnr commented Oct 28, 2021

Uh oh!

nikomatsakis left a comment

Choose a reason for hiding this comment

Uh oh!

nikomatsakisNov 4, 2021

Choose a reason for hiding this comment

Uh oh!

This comment has been minimized.

This comment has been minimized.

jackh726 commented Dec 4, 2021

Uh oh!

bors commented Dec 4, 2021

Uh oh!

Alexendoo commented Dec 6, 2021

Uh oh!

jackh726 commented Dec 7, 2021

Uh oh!

b-naber commented Dec 7, 2021

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

13 participants

b-naber commented Oct 18, 2021•
edited
Loading

lcnrOct 18, 2021•
edited
Loading