Allow const parameters in array sizes to be unified #60742

varkor · 2019-05-11T17:03:12Z

Fixes #60632.
Fixes #60744.
Fixes #60923.
(The last commit should probably be viewed in isolation, as it just renames things from type to kind.)

r? @eddyb

varkor · 2019-05-12T21:29:33Z

This changes some test output, but it's in very rare cases, and the messages are not incorrect (they're just a little extraneous).

bors · 2019-05-13T11:11:53Z

☔ The latest upstream changes (presumably #60765) made this pull request unmergeable. Please resolve the merge conflicts.

src/librustc/ty/relate.rs

eddyb · 2019-05-14T12:55:15Z

src/test/ui/consts/const-array-oob-arith.stderr

@@ -11,7 +11,7 @@ error[E0308]: mismatched types
  --> $DIR/const-array-oob-arith.rs:8:44
   |
 LL | const BOO: [i32; (ARR[0] - 41) as usize] = [5, 99];
-   |                                            ^^^^^^^ expected an array with a fixed size of 1 elements, found one with 2 elements
+   |                                            ^^^^^^^ expected `Const { ty: usize, val: Scalar(Bits { size: 8, bits: 1 }) }`, found `Const { ty: usize, val: Scalar(Bits { size: 8, bits: 2 }) }`


cc @oli-obk should we wait until printing this is fixed?

src/test/ui/const-generics/uninferred-consts-during-codegen-1.rs

varkor · 2019-05-14T21:08:16Z

Blocks #60839.
@bors p=1

yodaldevoid · 2019-05-18T18:41:58Z

I believe this is waiting on #59276, correct?

varkor · 2019-05-18T20:44:12Z

@yodaldevoid: that is correct.

varkor · 2019-05-20T22:03:45Z

I've added some tests to confirm this also fixes #60923.

varkor · 2019-05-25T20:01:09Z

@eddyb: I've updated the tests following #59276.

src/test/ui/consts/const-array-oob-arith.stderr

varkor · 2019-05-27T13:31:11Z

I've added back in the special cased message for arrays of different lengths and fixed a typo in related diagnostics. This is ready for re-review @eddyb.

eddyb · 2019-05-28T18:41:20Z

src/test/ui/const-generics/cannot-infer-type-for-const-param.stderr

@@ -14,7 +14,7 @@ error[E0308]: mismatched types
  --> $DIR/cannot-infer-type-for-const-param.rs:10:22
   |
 LL |     let _ = Foo::<3>([1, 2, 3]);
-   |                      ^^^^^^^^^ expected `Const { ty: usize, val: Unevaluated(DefId(0:18 ~ cannot_infer_type_for_const_param[317d]::main[0]::{{constant}}[0]), []) }`, found `Const { ty: usize, val: Scalar(0x0000000000000003) }`
+   |                      ^^^^^^^^^ expected `3`, found `3usize`


This... is really weird? I assume the 3 is @oli-obk's printing of same-crate anon consts?

I think it's because it can't infer the specific type, rather than an issue with printing. This is a bug, though, which should be fixed by #60839. (Which is ready to go as soon as this one is.)

I mean that the printing output is from Unevaluated for the 3 in Foo::<3> and not the value 3.

ugh, ok I did not envision user facing effects from that. So.... should I just do _ or {{unevaluated}} or sth?

I don't think we need to jump to something else here, but we should consider it carefully.
_ is clearly the conservative choice, as for "unevaluated" - I don't like that name too much, it just happened to be useful in ty::Const (and we could/should pick another name).

src/librustc/infer/resolve.rs

src/librustc/ty/error.rs

src/librustc/ty/relate.rs

… errors

varkor · 2019-05-28T20:35:44Z

@bors r=eddyb

bors · 2019-05-28T20:35:45Z

📌 Commit 6233d1f has been approved by eddyb

@eddyb

…eddyb Allow const parameters in array sizes to be unified Fixes rust-lang#60632. Fixes rust-lang#60744. Fixes rust-lang#60923. (The last commit should probably be viewed in isolation, as it just renames things from `type` to `kind`.) r? @eddyb

@eddyb

…eddyb Allow const parameters in array sizes to be unified Fixes rust-lang#60632. Fixes rust-lang#60744. Fixes rust-lang#60923. (The last commit should probably be viewed in isolation, as it just renames things from `type` to `kind`.) r? @eddyb

@ghost

Rollup of 9 pull requests Successful merges: - #60742 (Allow const parameters in array sizes to be unified) - #60756 (Add better tests for hidden lifetimes in impl trait) - #60928 (Changes the type `mir::Mir` into `mir::Body`) - #61024 (tests: Centralize proc macros commonly used for testing) - #61157 (BufReader: In Seek impl, remove extra discard_buffer call) - #61195 (Special-case `.llvm` in mangler) - #61202 (Print PermissionExt::mode() in octal in Documentation Examples) - #61259 (Mailmap fixes) - #61273 (mention that MaybeUninit is a bit like Option) Failed merges: r? @ghost

rust-highfive assigned eddyb May 11, 2019

rust-highfive added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label May 11, 2019

This was referenced May 11, 2019

Uninferred types/regions ICE for array impl with const generics #60744

Closed

Nightly compiler fails with ICE if const parameter is used to define type of function input #60632

Closed

This comment has been minimized.

Sign in to view

varkor mentioned this pull request May 12, 2019

Switch libcore array implementations to const generics. #60466

Closed

eddyb reviewed May 13, 2019

View reviewed changes

src/librustc/ty/relate.rs Outdated Show resolved Hide resolved

varkor force-pushed the fn-const-array-parameter branch 3 times, most recently from 7e94d17 to 6ced4f5 Compare May 13, 2019 20:32

eddyb reviewed May 14, 2019

View reviewed changes

src/test/ui/const-generics/uninferred-consts-during-codegen-1.rs Show resolved Hide resolved

davidtwco mentioned this pull request May 14, 2019

Fix ICE with struct ctors and const generics. #60839

Merged

davidtwco mentioned this pull request May 20, 2019

"Broken MIR" ICE when trying to use const generics #60923

Closed

This comment has been minimized.

Sign in to view

varkor force-pushed the fn-const-array-parameter branch 2 times, most recently from 83c32f5 to eee346f Compare May 25, 2019 19:42

bjorn3 reviewed May 26, 2019

View reviewed changes

src/test/ui/consts/const-array-oob-arith.stderr Outdated Show resolved Hide resolved

eddyb reviewed May 28, 2019

View reviewed changes

src/librustc/infer/resolve.rs Outdated Show resolved Hide resolved

eddyb reviewed May 28, 2019

View reviewed changes

src/librustc/ty/error.rs Show resolved Hide resolved

eddyb reviewed May 28, 2019

View reviewed changes

src/librustc/ty/relate.rs Outdated Show resolved Hide resolved

varkor added 15 commits May 28, 2019 21:34

Resolve consts in OpportunisticTypeResolver

4ad5c62

Add tests for uninferred consts during codegen

55dcc20

Rename OpportunisticTypeResolver to OpportunisticVarResolver

c94ba63

Update test output

908d97d

Eagerly evaluate in super_relate_consts

d5c6cb8

Remove FixedArraySize error

193b748

Add broken MIR regression tests

57ff589

Fix test after rebase

cfa1f80

Update tests after pretty printing

f865b7d

Use Display rather than Debug printing for const mismatch

f13317c

Reintroduce TypeError::FixedArraySize

8549953

Correct pluralisation of tuple/array/associated type binding mismatch…

56181cf

… errors

Make sure array length diagnostic doesn't regress

b3a13fd

Fix nits

2949160

Use assert_eq! instead of println! in tests

6233d1f

varkor force-pushed the fn-const-array-parameter branch from 5211145 to 6233d1f Compare May 28, 2019 20:35

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-author Status: This is awaiting some action (such as code changes or more information) from the author. labels May 28, 2019

Centril mentioned this pull request May 28, 2019

Rollup of 6 pull requests #61290

Closed

Centril mentioned this pull request May 28, 2019

Rollup of 9 pull requests #61296

Merged

bors merged commit 6233d1f into rust-lang:master May 29, 2019

yodaldevoid mentioned this pull request May 30, 2019

ICE on const generics in impl block ("failed to lift"). #60879

Closed

oli-obk mentioned this pull request Jun 6, 2019

Consider changing the way unevaluated constants are printed #61581

Open

varkor added the F-const_generics `#![feature(const_generics)]` label Oct 19, 2019

varkor mentioned this pull request Oct 19, 2019

Tracking issue for const generics (RFC 2000) #44580

Closed

16 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Allow const parameters in array sizes to be unified #60742

Allow const parameters in array sizes to be unified #60742

varkor commented May 11, 2019 •

edited

Loading

This comment has been minimized.

varkor commented May 12, 2019

bors commented May 13, 2019

eddyb May 14, 2019

varkor commented May 14, 2019

yodaldevoid commented May 18, 2019

varkor commented May 18, 2019

varkor commented May 20, 2019 •

edited

Loading

This comment has been minimized.

varkor commented May 25, 2019

varkor commented May 27, 2019

eddyb May 28, 2019

varkor May 28, 2019

eddyb May 28, 2019

oli-obk May 29, 2019

eddyb May 29, 2019

varkor commented May 28, 2019

bors commented May 28, 2019

Allow const parameters in array sizes to be unified #60742

Allow const parameters in array sizes to be unified #60742

Conversation

varkor commented May 11, 2019 • edited Loading

This comment has been minimized.

varkor commented May 12, 2019

bors commented May 13, 2019

eddyb May 14, 2019

Choose a reason for hiding this comment

varkor commented May 14, 2019

yodaldevoid commented May 18, 2019

varkor commented May 18, 2019

varkor commented May 20, 2019 • edited Loading

This comment has been minimized.

varkor commented May 25, 2019

varkor commented May 27, 2019

eddyb May 28, 2019

Choose a reason for hiding this comment

varkor May 28, 2019

Choose a reason for hiding this comment

eddyb May 28, 2019

Choose a reason for hiding this comment

oli-obk May 29, 2019

Choose a reason for hiding this comment

eddyb May 29, 2019

Choose a reason for hiding this comment

varkor commented May 28, 2019

bors commented May 28, 2019

varkor commented May 11, 2019 •

edited

Loading

varkor commented May 20, 2019 •

edited

Loading