De-abuse TyKind::Error in exhaustiveness checking #71930
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
(rust_highfive has picked a reviewer for you, use r? to override) |
rust-highfive
added
the
S-waiting-on-review
|
May I request a perf run on this? I'm not confident I managed to preserve performance |
|
@bors try @rust-timer queue |
|
Awaiting bors try build completion |
|
⌛ Trying commit e5667234a7fe2ca921766a197933229586749832 with merge 10bf6df2a72191317fadc1859f3d849707705018... |
|
The job Click to expand the log.I'm a bot! I can only do what humans tell me to, so if this was not helpful or you have suggestions for improvements, please ping or otherwise contact |
|
☀️ Try build successful - checks-actions, checks-azure |
|
Queued 10bf6df2a72191317fadc1859f3d849707705018 with parent f8d394e, future comparison URL. |
|
cc @eddyb (who I was working with on the tykind::error effort) |
8 tasks
|
It looks like the CI failure is a bug: |
mark-i-m
reviewed
May 6, 2020
mark-i-m
reviewed
May 6, 2020
|
Finished benchmarking try commit 10bf6df2a72191317fadc1859f3d849707705018, comparison URL. |
|
Perf looks neutral |
|
The regression was caused by 624ff54 . The regressed test is as follows: fn f() -> ! {
panic!("quux")
}
fn g() -> isize {
match f() {
true => 1,
false => 0,
}
}The reason it now fails is because instead of using the type of the first pattern to do our computations, we now use the type of the scrutinee. But that shouldn't make a difference, should it? Somehow here the scrutinee is given the type fn g() -> isize {
let x = f();
match x {
true => 1,
false => 0,
}
}So I don't think it's my fault; that looks like a type inference issue to me. For context, the type of the scrutinee is retrieved here |
|
|
Nadrieril
force-pushed
the
exhaustiveness-remove-tyerr
branch
from
e566723
to
a546f29
Compare
|
@matthewjasper Thanks, that was indeed the issue! |
|
The job Click to expand the log.I'm a bot! I can only do what humans tell me to, so if this was not helpful or you have suggestions for improvements, please ping or otherwise contact |
|
@matthewjasper Actually, your suggestion regressed this: fn from(never: !) -> TryFromIntError {
match never {}
}Apparently |
|
The job Click to expand the log.I'm a bot! I can only do what humans tell me to, so if this was not helpful or you have suggestions for improvements, please ping or otherwise contact |
|
|
Yeah I didn't mean to push that yet, I'm still hoping that @matthewjasper might pop in and figure out the arcane invocation needed to get the correct type x) |
Nadrieril
force-pushed
the
exhaustiveness-remove-tyerr
branch
from
19bfef8
to
c6f0947
Compare
|
@varkor I couldn't reproduce the perf impact locally, so I rebased on current master, but I still don't see it. On my machine I see a consistent ~1% improvement on the unicode_normalization tests :/ |
varkor
reviewed
May 18, 2020
There was a problem hiding this comment.
Thanks @Nadrieril, these comments are great :) I spotted one typo, but apart from that, I think this is ready to go.
|
I'm inclined to treat the perf results as negligible, especially if you see an improvement lately, but we'll make sure the PR isn't rolled up, so if it does have an effect, it'll be easier to track down. @bors rollup=never |
Co-authored-by: varkor <[email protected]>
|
Thanks! @bors r+ |
|
📌 Commit 4c6510b has been approved by |
bors
added
S-waiting-on-bors
RalfJung
reviewed
May 18, 2020
RalfJung
reviewed
May 18, 2020
|
@bors r=varkor |
|
📌 Commit d7e1d5f has been approved by |
|
🌲 The tree is currently closed for pull requests below priority 9000, this pull request will be tested once the tree is reopened |
|
☀️ Test successful - checks-azure |
eddyb
reviewed
May 23, 2020
bors
added a commit
to rust-lang-ci/rust
that referenced
this pull request
May 29, 2020
Exhaustiveness checking: work around type normalization issues This should resolve rust-lang#72476 and probably rust-lang#72467. This is a bit hacky but that's actually what the code was doing before rust-lang#71930. I'm essentially reverting rust-lang@e5a2cd5. So despite being hacky, it's been tried and tested (so much so that code relies on it now x)). Only the third commit does anything interesting.
Nadrieril
added
the
A-exhaustiveness-checking
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Replaces #71074. Context: #70866.
In order to remove the use of
TyKind::Error, I had to make sure we skip over those fields whose inhabitedness should not be observed. This is potentially error-prone however, since we must be careful not to mix filtered and unfiltered lists of patterns. I managed to hide away most of the filtering behind a newFieldsstruct, that I used everywhere relevant. I quite like the result; I think the twin concepts ofConstructorandFieldsmake a good mental model.As usual, I tried to separate commits that shuffle code around from commits that require more thought to review.
cc @varkor @Centril