[WIP] make LUB of higher-ranked things actually do EQ #44211

nikomatsakis · 2017-08-31T08:59:42Z

This is an experimental PR that I'm posting for the purposes of observing its effects via cargobomb (a crater run showed no regressions). The existing LUB/GLB code does some rather complex things around higher-ranked types (notably, grubbing about with region constraints) that I would like to move away from. I am not sure how often it is important in practice. This branch converts LUB/GLB of higher-ranked things into EQ.

r? @eddyb (this is just an experiment, don't really review =)

nikomatsakis · 2017-08-31T09:00:06Z

@bors try

bors · 2017-08-31T09:00:15Z

⌛ Trying commit 7226cda with merge ff1db21...

@eddyb

[DO NOT MERGE] make LUB of higher-ranked things actually do EQ This is an experimental PR that I'm posting for the purposes of observing its effects via cargobomb (a crater run showed [no regressions](https://gist.github.com/nikomatsakis/5aaf3782b4332a3ba408315cc84f69bf)). The existing LUB/GLB code does some rather complex things around higher-ranked types (notably, grubbing about with region constraints) that I would like to move away from. I am not sure how often it is important in practice. This branch converts LUB/GLB of higher-ranked things into EQ. r? @eddyb (this is just an experiment, don't really review =)

bors · 2017-08-31T10:35:04Z

☀️ Test successful - status-travis
State: approved= try=True

eddyb · 2017-08-31T11:02:18Z

cc @rust-lang/infra Cargobomb run requested.

Mark-Simulacrum · 2017-09-01T22:41:49Z

Marking as waiting on author since Travis failed. If these are intentional failures, please let us (infra team) know!

[01:07:43] running 23 tests
[01:07:43] ....FFFFFF.F.FFF.......
[01:07:43] failures:
[01:07:43]
[01:07:43] ---- test::glb_bound_free_infer stdout ----
[01:07:43]  thread 'test::glb_bound_free_infer' panicked at 'unexpected error computing LUB: RegionsInsufficientlyPolymorphic(BrAnon(1), '_#1r)', /checkout/src/librustc_driver/test.rs:418:22
[01:07:43] note: Run with `RUST_BACKTRACE=1` for a backtrace.
[01:07:43]
[01:07:43] ---- test::glb_bound_free stdout ----
[01:07:43]  thread 'test::glb_bound_free' panicked at 'unexpected error computing LUB: RegionsInsufficientlyPolymorphic(BrAnon(1), ReFree(DefId { krate: CrateNum(0), node: DefIndex(0) => test_crate/8cd878b }, BrAnon(1)))', /checkout/src/l
ibrustc_driver/test.rs:418:22
[01:07:43]
[01:07:43] ---- test::glb_bound_static stdout ----
[01:07:43]  thread 'test::glb_bound_static' panicked at 'unexpected error computing LUB: RegionsInsufficientlyPolymorphic(BrAnon(1), ReStatic)', /checkout/src/librustc_driver/test.rs:418:22
[01:07:43]
[01:07:43] ---- test::glb_free_free_with_common_scope stdout ----
[01:07:43]  thread 'test::glb_free_free_with_common_scope' panicked at 'Unexpected error: mismatched types Expected: []', /checkout/src/librustc_driver/test.rs:78:12
[01:07:43]
[01:07:43] ---- test::lub_bound_bound_inverse_order stdout ----
[01:07:43]  thread 'test::lub_bound_bound_inverse_order' panicked at 'unexpected error in LUB: expected bound lifetime parameterBrAnon(2), found concrete lifetime', /checkout/src/librustc_driver/test.rs:410:26
[01:07:43]
[01:07:43] ---- test::lub_bound_free stdout ----
[01:07:43]  thread 'test::lub_bound_free' panicked at 'unexpected error in LUB: expected bound lifetime parameterBrAnon(1), found concrete lifetime', /checkout/src/librustc_driver/test.rs:410:26
[01:07:43]
[01:07:43] ---- test::lub_bound_static stdout ----
[01:07:43]  thread 'test::lub_bound_static' panicked at 'unexpected error in LUB: expected bound lifetime parameterBrAnon(1), found concrete lifetime', /checkout/src/librustc_driver/test.rs:410:26
[01:07:43]
[01:07:43] ---- test::lub_free_bound_infer stdout ----
[01:07:43]  thread 'test::lub_free_bound_infer' panicked at 'unexpected error in LUB: expected bound lifetime parameterBrAnon(1), found concrete lifetime', /checkout/src/librustc_driver/test.rs:410:26
[01:07:43]
[01:07:43] ---- test::lub_returning_scope stdout ----
[01:07:43]  thread 'test::lub_returning_scope' panicked at 'Unexpected error: mismatched types Expected: []', /checkout/src/librustc_driver/test.rs:78:12
[01:07:43]
[01:07:43] ---- test::lub_free_free stdout ----
[01:07:43]  thread 'test::lub_free_free' panicked at 'Unexpected error: mismatched types Expected: []', /checkout/src/librustc_driver/test.rs:78:12
[01:07:43]
[01:07:43]
[01:07:43] failures:
[01:07:43]     test::glb_bound_free
[01:07:43]     test::glb_bound_free_infer
[01:07:43]     test::glb_bound_static
[01:07:43]     test::glb_free_free_with_common_scope
[01:07:43]     test::lub_bound_bound_inverse_order
[01:07:43]     test::lub_bound_free
[01:07:43]     test::lub_bound_static
[01:07:43]     test::lub_free_bound_infer
[01:07:43]     test::lub_free_free
[01:07:43]     test::lub_returning_scope

eddyb · 2017-09-01T22:59:34Z

@Mark-Simulacrum This is not supposed to be merged and those test failures probably make sense given the changes - we only need a cargobomb run result from it.

aidanhs · 2017-09-05T20:26:11Z

Cargobomb run started. Sorry for the delay, we didn't have spare boxes and then I was away for a few days.

Mark-Simulacrum · 2017-09-09T23:31:39Z

Run is complete: http://cargobomb-reports.s3.amazonaws.com/pr-44211/index.html. 15 regressions straight off, but haven't triaged.

eddyb · 2017-09-11T05:38:35Z

@nikomatsakis Looks like no actual regressions to me, only some flaky tests and timeouts.

arielb1 · 2017-09-18T21:04:22Z

Given that there were no regressions of crater, I think we can merge this.

However, I would like it if on an LUB failure, we could at least have a special error message noting the new compiler update.

eddyb · 2017-09-18T21:44:09Z

@arielb1 from what @nikomatsakis was telling me, this branch is only an experiment to confirm that a separate reimplementation strategy could work without breaking existing code.

nikomatsakis · 2017-09-20T16:42:31Z

While @eddyb is correct, now that the experiment was successful, I've been contemplating landing (some variant of) this branch, just to pave the way for the future work. It should be possible to make a variant of this that offers a specialized error message, at least.

aidanhs · 2017-09-28T01:02:58Z

@nikomatsakis just pinging you for triage purposes to see how your contemplation of this PR mentioned in your last comment is going :)

aidanhs · 2017-10-05T19:08:19Z

@nikomatsakis ping again to see where this PR is up to! If this was just an experimental PR, maybe worth closing and reopening a fresh one with changes to be reviewed/merged?

nikomatsakis · 2017-10-09T20:31:01Z

@aidanhs I think we've decided to turn into something a bit more than experimental. I'm still hoping to refresh the approach in the next day or two, so I'm inclined not to close just yet.

This reverts commit 7226cda.

carols10cents · 2017-10-16T15:11:17Z

I'm still hoping to refresh the approach in the next day or two, so I'm inclined not to close just yet.

7 days ago

Triage ping! How's it going?

nikomatsakis · 2017-10-16T17:43:25Z

Some progress, got blocked by other things.

shepmaster · 2017-11-03T13:37:00Z

This has been inactive for 18 days, so I'm going to go ahead and close this to keep things tidy. As always, please reopen once some updates happen! Thanks!

@arielb1

Simplify higher-ranked LUB/GLB This is a better version of #44211. It still makes higher-ranked LUB/GLB into a hard equality test, however, it does try to identify that something changed and issue a notice to the user. I wroteup #45852 as a tracking issue for this change. Currently, this moves straight to a hard-error, on the basis that the crater run in #44211 saw no impact. It might be good to retest -- or perhaps to try for a warning period. Trying to do the latter in a precise way would be somewhat painful, but an imprecise way might suffice -- that is, we could issue warning *whenever* a LUB/GLB operation succeeds that will later fail, even if it doesn't ultimately impact the type check. I could experiment with this. ~~I am *mildly* wary about landing this independently of other code that moves to a universe-based system. In particular, I was nervous that this change would make coherence accepts new pairs of impls that will later be errors. I have the code for the universe-based approach available, I hope to open an PR and run some tests on its impact very shortly.~~ @arielb1 points out that I was being silly. r? @arielb1

@arielb1

Simplify higher-ranked LUB/GLB This is a better version of #44211. It still makes higher-ranked LUB/GLB into a hard equality test, however, it does try to identify that something changed and issue a notice to the user. I wroteup #45852 as a tracking issue for this change. Currently, this moves straight to a hard-error, on the basis that the crater run in #44211 saw no impact. It might be good to retest -- or perhaps to try for a warning period. Trying to do the latter in a precise way would be somewhat painful, but an imprecise way might suffice -- that is, we could issue warning *whenever* a LUB/GLB operation succeeds that will later fail, even if it doesn't ultimately impact the type check. I could experiment with this. ~~I am *mildly* wary about landing this independently of other code that moves to a universe-based system. In particular, I was nervous that this change would make coherence accepts new pairs of impls that will later be errors. I have the code for the universe-based approach available, I hope to open an PR and run some tests on its impact very shortly.~~ @arielb1 points out that I was being silly. r? @arielb1

@arielb1

Simplify higher-ranked LUB/GLB This is a better version of #44211. It still makes higher-ranked LUB/GLB into a hard equality test, however, it does try to identify that something changed and issue a notice to the user. I wroteup #45852 as a tracking issue for this change. Currently, this moves straight to a hard-error, on the basis that the crater run in #44211 saw no impact. It might be good to retest -- or perhaps to try for a warning period. Trying to do the latter in a precise way would be somewhat painful, but an imprecise way might suffice -- that is, we could issue warning *whenever* a LUB/GLB operation succeeds that will later fail, even if it doesn't ultimately impact the type check. I could experiment with this. ~~I am *mildly* wary about landing this independently of other code that moves to a universe-based system. In particular, I was nervous that this change would make coherence accepts new pairs of impls that will later be errors. I have the code for the universe-based approach available, I hope to open an PR and run some tests on its impact very shortly.~~ @arielb1 points out that I was being silly. r? @arielb1

rust-highfive assigned eddyb Aug 31, 2017

aidanhs added the S-waiting-on-crater Status: Waiting on a crater run to be completed. label Aug 31, 2017

Mark-Simulacrum added S-waiting-on-crater Status: Waiting on a crater run to be completed. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Sep 2, 2017

eddyb mentioned this pull request Sep 11, 2017

Blacklist update needed rust-lang/crater#126

Closed

Mark-Simulacrum added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. and removed S-waiting-on-crater Status: Waiting on a crater run to be completed. labels Sep 12, 2017

aidanhs mentioned this pull request Sep 28, 2017

Add a DynSized trait #44469

Closed

nikomatsakis added 2 commits October 9, 2017 16:37

make LUB/GLB of higher-ranked things actually do EQ

75012f4

Revert "make LUB/GLB of higher-ranked things actually do EQ"

083c433

This reverts commit 7226cda.

nikomatsakis force-pushed the chalk-simplify-hr-lub-glb branch from 7226cda to 083c433 Compare October 9, 2017 20:37

nikomatsakis changed the title ~~[DO NOT MERGE] make LUB of higher-ranked things actually do EQ~~ [WIP] make LUB of higher-ranked things actually do EQ Oct 9, 2017

shepmaster closed this Nov 3, 2017

nikomatsakis mentioned this pull request Nov 7, 2017

Simplify higher-ranked LUB/GLB #45853

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[WIP] make LUB of higher-ranked things actually do EQ #44211

[WIP] make LUB of higher-ranked things actually do EQ #44211

nikomatsakis commented Aug 31, 2017

nikomatsakis commented Aug 31, 2017

bors commented Aug 31, 2017

bors commented Aug 31, 2017

eddyb commented Aug 31, 2017

Mark-Simulacrum commented Sep 1, 2017

eddyb commented Sep 1, 2017

aidanhs commented Sep 5, 2017

Mark-Simulacrum commented Sep 9, 2017

eddyb commented Sep 11, 2017

arielb1 commented Sep 18, 2017 •

edited

Loading

eddyb commented Sep 18, 2017

nikomatsakis commented Sep 20, 2017

aidanhs commented Sep 28, 2017

aidanhs commented Oct 5, 2017

nikomatsakis commented Oct 9, 2017

carols10cents commented Oct 16, 2017

nikomatsakis commented Oct 16, 2017

shepmaster commented Nov 3, 2017

[WIP] make LUB of higher-ranked things actually do EQ #44211

[WIP] make LUB of higher-ranked things actually do EQ #44211

Conversation

nikomatsakis commented Aug 31, 2017

nikomatsakis commented Aug 31, 2017

bors commented Aug 31, 2017

bors commented Aug 31, 2017

eddyb commented Aug 31, 2017

Mark-Simulacrum commented Sep 1, 2017

eddyb commented Sep 1, 2017

aidanhs commented Sep 5, 2017

Mark-Simulacrum commented Sep 9, 2017

eddyb commented Sep 11, 2017

arielb1 commented Sep 18, 2017 • edited Loading

eddyb commented Sep 18, 2017

nikomatsakis commented Sep 20, 2017

aidanhs commented Sep 28, 2017

aidanhs commented Oct 5, 2017

nikomatsakis commented Oct 9, 2017

carols10cents commented Oct 16, 2017

nikomatsakis commented Oct 16, 2017

shepmaster commented Nov 3, 2017

arielb1 commented Sep 18, 2017 •

edited

Loading