Normalize all opaque types when converting ParamEnv to Reveal::All by Aaron1011 · Pull Request #65989 · rust-lang/rust

Aaron1011 · 2019-10-31T01:24:19Z

When we normalize a type using a ParamEnv with a reveal mode of
RevealMode::All, we will normalize opaque types to their underlying
types (e.g. type MyOpaque = impl Foo -> StructThatImplsFoo).
However, the ParamEnv may still have predicates referring to the
un-normalized opaque type (e.g. <T as MyTrait<MyOpaque>>). This can
cause trait projection to fail, since a type containing normalized
opaque types will not match up with the un-normalized type in the
ParamEnv.

To fix this, we now explicitly normalize all opaque types in
caller_bounds of a ParamEnv when changing its mode to
RevealMode::All. This ensures that all predicatse will refer to the
underlying types of any opaque types involved, allowing them to be
matched up properly during projection. To reflect the fact that
normalization is occuring, ParamEnv::with_reveal_all is renamed to
ParamEnv::with_reveal_all_normalized

Fixes#65918

rust-highfive · 2019-10-31T01:24:23Z

r? @petrochenkov

(rust_highfive has picked a reviewer for you, use r? to override)

Centril

Just some nits. :)
Nice PR!

Aaron1011 · 2019-10-31T02:24:28Z

@Centril I've addressed your comments.

petrochenkov · 2019-11-01T21:11:48Z

r? @eddyb for review or reassignment

eddyb · 2019-11-02T05:45:03Z

This is @rust-lang/wg-traits territory.
r? @nikomatsakis

JohnCSimon · 2019-11-09T07:17:31Z

Ping from triage:
@nikomatsakis can you please review this?
cc: @petrochenkov @Aaron1011 @petrochenkov

Thanks!

JohnCSimon · 2019-11-16T23:56:27Z

Pinging again from triage:
@nikomatsakis can you please review this?
cc: @petrochenkov @Aaron1011 @petrochenkov @rust-lang/wg-traits @nikomatsakis

Thanks!

Dylan-DPC-zz · 2019-11-24T13:43:32Z

r? @petrochenkov

petrochenkov · 2019-11-24T17:25:31Z

I'm not qualified to review this.
r? @nikomatsakis

bors · 2019-12-01T03:39:43Z

☔ The latest upstream changes (presumably #66908) made this pull request unmergeable. Please resolve the merge conflicts.

nikomatsakis · 2019-12-07T00:34:16Z

Will review -- sorry for radio silence

nikomatsakis

Yet another example of why it would be good to pursue an alternative normalization strategy, I suppose. I am curious though whether this has any performance impact. We may want to think about more aggressive caching.

nikomatsakis · 2019-12-07T00:38:35Z

@bors try

bors · 2019-12-07T00:38:59Z

⌛ Trying commit 10a5c93 with merge ff0aa14...

Normalize all opaque types when converting ParamEnv to Reveal::All When we normalize a type using a ParamEnv with a reveal mode of RevealMode::All, we will normalize opaque types to their underlying types (e.g. `type MyOpaque = impl Foo` -> `StructThatImplsFoo`). However, the ParamEnv may still have predicates referring to the un-normalized opaque type (e.g. `<T as MyTrait<MyOpaque>>`). This can cause trait projection to fail, since a type containing normalized opaque types will not match up with the un-normalized type in the `ParamEnv`. To fix this, we now explicitly normalize all opaque types in caller_bounds of a `ParamEnv` when changing its mode to `RevealMode::All`. This ensures that all predicatse will refer to the underlying types of any opaque types involved, allowing them to be matched up properly during projection. To reflect the fact that normalization is occuring, `ParamEnv::with_reveal_all` is renamed to `ParamEnv::with_reveal_all_normalized` Fixes#65918

bors · 2019-12-07T03:20:30Z

☀️ Try build successful - checks-azure
Build commit: ff0aa14 (ff0aa146df927e3857e94e7206d788076f3b500b)

Aaron1011 · 2019-12-08T17:05:30Z

@nikomatsakis: Did you mean to do a perf run as well as a try build?

nikomatsakis · 2019-12-10T22:03:03Z

@Aaron1011 dang it, yes I did

bors · 2020-06-13T03:48:23Z

☀️ Try build successful - checks-azure
Build commit: cd9893430843101b421d45480d3dcb55c4dce64f (cd9893430843101b421d45480d3dcb55c4dce64f)

rust-timer · 2020-06-13T03:48:25Z

Queued cd9893430843101b421d45480d3dcb55c4dce64f with parent f4fbb93, future comparison URL.

rust-timer · 2020-06-13T06:56:42Z

Finished benchmarking try commit (cd9893430843101b421d45480d3dcb55c4dce64f): comparison url.

nikomatsakis · 2020-06-15T18:06:27Z

Well, it definitely hurts the stress tests =)

Dylan-DPC-zz · 2020-07-18T01:12:26Z

@Aaron1011 any updates on this?

nikomatsakis · 2020-07-20T13:55:37Z

@matthewjasper was working on a PR that improved the perf of the stress tests quite a bit, might be worth retesting. I'm not sure though if that PR landed yet.

Fixesrust-lang#65918

Aaron1011 · 2020-07-22T06:41:49Z

@bors try @rust-timer queue

rust-timer · 2020-07-22T06:41:51Z

Awaiting bors try build completion

bors · 2020-07-22T06:42:02Z

⌛ Trying commit 5e2e927 with merge 95dcfa37cf97c1d630a7820e31d39a7a37ae2c1d...

bors · 2020-07-22T08:49:07Z

☀️ Try build successful - checks-actions, checks-azure
Build commit: 95dcfa37cf97c1d630a7820e31d39a7a37ae2c1d (95dcfa37cf97c1d630a7820e31d39a7a37ae2c1d)

rust-timer · 2020-07-22T08:49:09Z

Queued 95dcfa37cf97c1d630a7820e31d39a7a37ae2c1d with parent 4825e12, future comparison URL.

rust-timer · 2020-07-22T15:54:15Z

Finished benchmarking try commit (95dcfa37cf97c1d630a7820e31d39a7a37ae2c1d): comparison url.

Benchmarking this pull request likely means that it is perf-sensitive, so we're automatically marking it as not fit for rolling up. Please note that if the perf results are neutral, you should likely undo the rollup=never given below by specifying rollup- to bors.

Importantly, though, if the results of this run are non-neutral do not roll this PR up -- it will mask other regressions or improvements in the roll up.

@bors rollup=never

Aaron1011 · 2020-07-22T19:02:20Z

@nikomatsakis: It looks like almost all of the perf impact is gone.

nikomatsakis · 2020-07-27T17:19:49Z

@bors r+

bors · 2020-07-27T17:19:50Z

📌 Commit 5e2e927 has been approved by nikomatsakis

bors · 2020-07-27T18:12:57Z

⌛ Testing commit 5e2e927 with merge a7b1b8ea5d0b7b7e6d162fe0066eb0f1afc94778...

bors · 2020-07-27T18:42:46Z

💔 Test failed - checks-actions

Aaron1011 · 2020-07-31T13:39:08Z

I think this was a spurious failure.

@bors r=nikomatsakis

bors · 2020-07-31T13:39:09Z

💡 This pull request was already approved, no need to approve it again.

This pull request previously failed. You should add more commits to fix the bug, or use retry to trigger a build again.
There's another pull request that is currently being tested, blocking this pull request: Enable the profiler on FreeBSD #74844

bors · 2020-07-31T13:39:10Z

📌 Commit 5e2e927 has been approved by nikomatsakis

bors · 2020-07-31T18:15:10Z

⌛ Testing commit 5e2e927 with merge 6e87bac...

bors · 2020-07-31T20:08:02Z

☀️ Test successful - checks-actions, checks-azure
Approved by: nikomatsakis
Pushing 6e87bac to master...

Mark-Simulacrum · 2020-08-03T22:14:04Z

This was a slight performance regression, as expected.

rust-highfive assigned petrochenkov Oct 31, 2019

rust-highfive added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Oct 31, 2019

Centril added the F-type_alias_impl_trait `#[feature(type_alias_impl_trait)]` label Oct 31, 2019

Centril reviewed Oct 31, 2019
View reviewed changes

rust-highfive assigned eddyb and unassigned petrochenkov Nov 1, 2019

rust-highfive assigned nikomatsakis and unassigned eddyb Nov 2, 2019

rust-highfive assigned petrochenkov and unassigned nikomatsakis Nov 24, 2019

rust-highfive assigned nikomatsakis and unassigned petrochenkov Nov 24, 2019

Aaron1011 force-pushed the fix/normalize-param-env branch from 6d6b5d5 to 33e196dCompare December 2, 2019 17:58

nikomatsakis reviewed Dec 7, 2019
View reviewed changes

Aaron1011 added 4 commits July 22, 2020 00:57

Normalize opaque types when converting ParamEnv to Reveal::All
ac9dfc3
Fixesrust-lang#65918

Erase regions in try_eval_bits
117a60e

Skip computing param_env and size if not needed
90aee14

Fix rebase fallout
5e2e927

nikomatsakis approved these changes Jul 27, 2020
View reviewed changes

This was referenced Jul 31, 2020
Rename HAIR to THIR (Typed HIR). #74582
Merged
Rollup of 4 pull requests #74987
Closed

Uh oh!

Uh oh!

Conversation

Aaron1011 commented Oct 31, 2019

Uh oh!

rust-highfive commented Oct 31, 2019

Uh oh!

Centril left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Aaron1011 commented Oct 31, 2019

Uh oh!

petrochenkov commented Nov 1, 2019

Uh oh!

eddyb commented Nov 2, 2019

Uh oh!

JohnCSimon commented Nov 9, 2019

Uh oh!

JohnCSimon commented Nov 16, 2019

Uh oh!

Dylan-DPC-zz commented Nov 24, 2019

Uh oh!

petrochenkov commented Nov 24, 2019

Uh oh!

bors commented Dec 1, 2019

Uh oh!

nikomatsakis commented Dec 7, 2019

Uh oh!

nikomatsakis left a comment

Choose a reason for hiding this comment

Uh oh!

nikomatsakis commented Dec 7, 2019

Uh oh!

bors commented Dec 7, 2019

Uh oh!

bors commented Dec 7, 2019

Uh oh!

Aaron1011 commented Dec 8, 2019

Uh oh!

nikomatsakis commented Dec 10, 2019

Uh oh!

bors commented Jun 13, 2020

Uh oh!

rust-timer commented Jun 13, 2020

Uh oh!

rust-timer commented Jun 13, 2020

Uh oh!

nikomatsakis commented Jun 15, 2020

Uh oh!

Dylan-DPC-zz commented Jul 18, 2020

Uh oh!

nikomatsakis commented Jul 20, 2020

Uh oh!

Aaron1011 commented Jul 22, 2020

Uh oh!

rust-timer commented Jul 22, 2020

Uh oh!

bors commented Jul 22, 2020

Uh oh!

bors commented Jul 22, 2020

Uh oh!

rust-timer commented Jul 22, 2020

Uh oh!

rust-timer commented Jul 22, 2020

Uh oh!

Aaron1011 commented Jul 22, 2020• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

nikomatsakis commented Jul 27, 2020

Uh oh!

bors commented Jul 27, 2020

Uh oh!

Aaron1011 commented Jul 22, 2020•
edited
Loading