Rustup to rustc 1.41.0-nightly (35ef33a8 2019-11-21) #4825

Manishearth · 2019-11-18T15:51:38Z

I don't have the right fix for the fmtstr tests, and I'm also hitting problems caused by messense/rustc-test#3

List of rustups:

syntax: Keep string literals in ABIs and asm! more precisely rust#66271 (syntax: Keep string literals in ABIs and asm! more precisely)
Stabilize ! in Rust 1.41.0 rust#65355 (Stabilize ! in Rust 1.41.0)
Reduce size of hir::Expr by boxing more of hir::InlineAsm rust#66515 (Reduce size of hir::Expr by boxing more of hir::InlineAsm)
Specific labels when referring to "expected" and "found" types rust#66389 (Specific labels when referring to "expected" and "found" types)
[mir-opt] Turn on the ConstProp pass by default rust#66074 ([mir-opt] Turn on the ConstProp pass by default)

changelog: none

Manishearth · 2019-11-18T16:05:51Z

Working on getting the compiletest stuff fixed, does someone want to look at the write/print test failures?

flip1995 · 2019-11-18T16:16:12Z

I did pretty much the same as this PR, but was stuck at the tester thing in compiletest.

Do we really want to depend on tester? I think, we should keep hiding it behind the #[cfg(not(rustc))] in compiletest_rs.

cc Manishearth/compiletest-rs#203 (comment)

Manishearth · 2019-11-18T16:26:49Z

Using the rustc feature caused similar issues, and it would be nice to never be broken by a libtest change again.

Manishearth · 2019-11-18T18:00:54Z

If we can land this with the rustc feature that's fine too. But we still have an ICE and test failures on the print/write tests.

They may have to do with the difference between litstr.symbol and litstr.symbol_unescaped

flip1995 · 2019-11-18T22:49:54Z

Both the rustc feature and the tester dep require a change in compiletest or tester respectively, so we can just go for tester, if that will be the future way to go with compiletest.

What test failures? All the print/write ui-tests are green. What am I missing?

flip1995 · 2019-11-18T22:51:47Z

require a change in compiletest or tester respectively

except we just ignore the failing ui-test for now, of course.

Manishearth · 2019-11-18T22:52:09Z

Hmm, I was getting test failures locally with this branch

flip1995 · 2019-11-18T23:03:22Z

Should we r+ this with module_name_repetitions test ignored and fix it later, once tester is renamed?

Hmm, I was getting test failures locally with this branch

Weird... I didn't and travis is also green 🤔

Manishearth · 2019-11-18T23:07:14Z

Sure, go for it.

…

On Mon, Nov 18, 2019, 3:03 PM Philipp Krones ***@***.***> wrote: Should we r+ this with module_name_repetitions test ignored and fix it later, once tester is renamed? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#4825?email_source=notifications&email_token=AAMK6SAQHGOGTKIMQ3OGMM3QUMNL3A5CNFSM4JOVRPKKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEEMHKPI#issuecomment-555251005>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAMK6SC7ZWADXF2COVAYQH3QUMNL3ANCNFSM4JOVRPKA> .

flip1995 · 2019-11-18T23:13:23Z

I'm only on mobile right now. I'll get back to this tomorrow (if no one else did it before that)

Manishearth · 2019-11-19T07:52:34Z

@bors r=flip1995

turns out there were no write failures 🤷‍♂️

bors · 2019-11-19T07:52:35Z

📌 Commit aa5a95f has been approved by flip1995

Rustup to rustc 1.41.0-nightly (a0d40f8bd 2019-11-18) I don't have the right fix for the fmtstr tests, and I'm also hitting problems caused by messense/rustc-test#3 changelog: none

bors · 2019-11-19T07:52:47Z

⌛ Testing commit aa5a95f with merge 37a8749...

bors · 2019-11-19T08:05:19Z

💔 Test failed - status-appveyor

Manishearth · 2019-11-19T08:33:40Z

New failures have come in, fixed one of them but redundant_clone is still failing and I'm going to bed for now.

flip1995 · 2019-11-19T11:56:54Z

tests/ui/missing_const_for_fn/cant_be_const.rs

@@ -32,11 +32,8 @@ fn random_caller() -> u32 {

 static Y: u32 = 0;

-// We should not suggest to make this function `const` because const functions are not allowed to
-// refer to a static variable
 fn get_y() -> u32 {


This shouldn't be linted still. Adding const to this function gives the error described in the comment. But for some reason rustc_mir::transform::qualify_min_const_fn::is_min_const_fn returns Ok(()) for this function:

rust-clippy/clippy_lints/src/missing_const_for_fn.rs

Lines 108 to 110 in b4f1769

let mir = cx.tcx.optimized_mir(def_id);

if let Err((span, err)) = is_min_const_fn(cx.tcx, def_id, &mir) {

cc @oli-obk Does optimized_mir replace the static with a const value and is_min_const_fn then thinks this could be a const fn?

Did something change in the handling of Statements in MIR recently? The error in redundant_clone is similar. This code

rust-clippy/clippy_lints/src/redundant_clone.rs

Lines 374 to 383 in b4f1769

fn visit_local(&mut self, local: &mir::Local, ctx: PlaceContext, _: mir::Location) {

match ctx {

PlaceContext::MutatingUse(MutatingUseContext::Drop) | PlaceContext::NonUse(_) => return,

_ => {},

}

if *local == self.local {

self.used_other_than_drop = true;

}

}

stopped detecting, that in (a.clone(), a) both as are the same.

What I mean with

The error [...] is similar

is, that in both cases the code that should check for them is never actually reached by the traversal of the Statement (or the Terminator?)

oh... we run is_min_const_fn on the optimized mir while rustc runs it on mir_const. This will keep causing problems. One thing clippy could do is to make the optimized_mir query not do any optimizations. This is possible since rust-lang/rust#66297

So how do I override the optimized_mir query? When I just copy and paste its source code [1] from rustc and remove the run_optimization_passes call¹, a lot of tests began to ICE. I added this:

config.override_queries = Some(|_, providers, _| { providers.optimized_mir = |tcx, def_id| { // (Mir-)Borrowck uses `mir_validated`, so we have to force it to // execute before we can steal. tcx.ensure().mir_borrowck(def_id); let (body, _) = tcx.mir_validated(def_id); let body = body.steal(); tcx.arena.alloc(body) } });

to

rust-clippy/src/driver.rs

Lines 63 to 64 in b4f1769

impl rustc_driver::Callbacks for ClippyCallbacks {

fn config(&mut self, config: &mut interface::Config) {

¹ I also removed this:

if tcx.is_constructor(def_id) { // There's no reason to run all of the MIR passes on constructors when // we can just output the MIR we want directly. This also saves const // qualification and borrow checking the trouble of special casing // constructors. return shim::build_adt_ctor(tcx, def_id); }

because the shim module is private.

[1] https://github.com/rust-lang/rust/blob/9d6ff1553b7debbe5c99c102ce0978b6130592f8/src/librustc_mir/transform/mod.rs#L316-L333

Hmm. So we'd need to make more things public. But that's probably not the failure reason. One thing you can try is set the mir opt level to 0 instead of trying to replace the query. If that doesn't work, reintroduce the mir optimizations but remove all the ones that break clippy (e.g. const_prop)

One thing you can try is set the mir opt level to 0

How do I set the opt level? And how can I enable/disable mir passes?

Stablized in rust-lang/rust#65355

Fixes rust-lang#4835

bors · 2019-11-23T00:31:22Z

⌛ Testing commit 7bae5bd with merge 19769b5...

Rustup to rustc 1.41.0-nightly (35ef33a8 2019-11-21) I don't have the right fix for the fmtstr tests, and I'm also hitting problems caused by messense/rustc-test#3 List of rustups: - rust-lang/rust#66271 (syntax: Keep string literals in ABIs and `asm!` more precisely) - rust-lang/rust#65355 (Stabilize `!` in Rust 1.41.0) - rust-lang/rust#66515 (Reduce size of `hir::Expr` by boxing more of `hir::InlineAsm`) - rust-lang/rust#66389 (Specific labels when referring to "expected" and "found" types) - rust-lang/rust#66074 ([mir-opt] Turn on the `ConstProp` pass by default) changelog: none

bors · 2019-11-23T00:49:51Z

💔 Test failed - checks-travis

flip1995 · 2019-11-23T01:16:52Z

@bors retry

Rustup to rustc 1.41.0-nightly (35ef33a8 2019-11-21) I don't have the right fix for the fmtstr tests, and I'm also hitting problems caused by messense/rustc-test#3 List of rustups: - rust-lang/rust#66271 (syntax: Keep string literals in ABIs and `asm!` more precisely) - rust-lang/rust#65355 (Stabilize `!` in Rust 1.41.0) - rust-lang/rust#66515 (Reduce size of `hir::Expr` by boxing more of `hir::InlineAsm`) - rust-lang/rust#66389 (Specific labels when referring to "expected" and "found" types) - rust-lang/rust#66074 ([mir-opt] Turn on the `ConstProp` pass by default) changelog: none

bors · 2019-11-23T01:17:00Z

⌛ Testing commit 7bae5bd with merge 3870050...

bors · 2019-11-23T01:18:04Z

💔 Test failed - status-appveyor

flip1995 · 2019-11-23T10:58:36Z

uh that is bad: https://travis-ci.com/rust-lang/rust-clippy/jobs/259722402#L5480-L5487

Since when do we lint the output of build scripts? 😮

flip1995 · 2019-11-23T11:08:38Z

This is even worse: https://travis-ci.com/rust-lang/rust-clippy/jobs/259722402#L7287

The match_same_arms lint triggers once for every combination of identical match arms, so

match a {
    1 => (),
    2 => (),
    3 => (),
    _ => (),
}

would trigger 6 times: 2<->1, 3<->1, 3<->2, _<->1, _<->2, _<->3. But that's an issue for another day.

flip1995 · 2019-11-23T12:32:14Z

Since when do we lint the output of build scripts?

It seems since ever. We probably shouldn't dump the Clippy output to the travis terminal though.

[WIP] RIIR: Integration tests In #4825 the `rust-lang/chalk` test failed because the output was too large. I didn't want to completely disabling the output, since showing the backtrace of an ICE directly in travis is pretty useful. Since finding strings in command outputs is easier in Rust, than in bash, I just RIIRed it. This and also rewriting our tests in Rust may help with trying out new CI platforms (cc #4577) Based on #4825, so I could test it. changelog: none

ICEs because of rustc, not Clippy

flip1995 · 2019-11-23T15:54:29Z

@bors r+

bors · 2019-11-23T15:54:30Z

📌 Commit 553db87 has been approved by flip1995

bors · 2019-11-23T15:54:37Z

⌛ Testing commit 553db87 with merge 60e8413...

Rustup to rustc 1.41.0-nightly (35ef33a8 2019-11-21) I don't have the right fix for the fmtstr tests, and I'm also hitting problems caused by messense/rustc-test#3 List of rustups: - rust-lang/rust#66271 (syntax: Keep string literals in ABIs and `asm!` more precisely) - rust-lang/rust#65355 (Stabilize `!` in Rust 1.41.0) - rust-lang/rust#66515 (Reduce size of `hir::Expr` by boxing more of `hir::InlineAsm`) - rust-lang/rust#66389 (Specific labels when referring to "expected" and "found" types) - rust-lang/rust#66074 ([mir-opt] Turn on the `ConstProp` pass by default) changelog: none

bors · 2019-11-23T16:14:49Z

☀️ Test successful - checks-travis, status-appveyor
Approved by: flip1995
Pushing 60e8413 to master...

[WIP] RIIR: Integration tests In #4825 the `rust-lang/chalk` test failed because the output was too large. I didn't want to completely disabling the output, since showing the backtrace of an ICE directly in travis is pretty useful. Since finding strings in command outputs is easier in Rust, than in bash, I just RIIRed it. This and also rewriting our tests in Rust may help with trying out new CI platforms (cc #4577) changelog: none

RIIR: Integration tests In #4825 the `rust-lang/chalk` test failed because the output was too large. I didn't want to completely disabling the output, since showing the backtrace of an ICE directly in travis is pretty useful. Since finding strings in command outputs is easier in Rust, than in bash, I just RIIRed it. This and also rewriting our tests in Rust may help with trying out new CI platforms (cc #4577) changelog: none

Fix redundant_clone lint not working with PathBuf and OsString #4825 diabled MIR optimization in clippy, including `rustc_mir::transform::InstCombine` which reduces `&(*x)` to `x`. This PR tries to unwrap `&*` when looking into `mir::Rvalue`s. Fixes #5014. --- changelog: fixed `redundant_clone` lint not working with `PathBuf` and `OsString`

This comment has been minimized.

Sign in to view

Manishearth force-pushed the rustup branch from f21db0a to 9325a85 Compare November 19, 2019 07:27

Rustup to rustc 1.41.0-nightly (a0d40f8bd 2019-11-18)

aa5a95f

Manishearth force-pushed the rustup branch from 9325a85 to aa5a95f Compare November 19, 2019 07:50

Manishearth changed the title ~~[WIP] Rustup to rustc 1.41.0-nightly (a0d40f8bd 2019-11-18)~~ Rustup to rustc 1.41.0-nightly (a0d40f8bd 2019-11-18) Nov 19, 2019

Rustup to rustc 1.41.0-nightly (d1da8023d 2019-11-19)

d183bda

flip1995 reviewed Nov 19, 2019

View reviewed changes

This was referenced Nov 19, 2019

New lint: Implement ifs_same_cond_fn #4814

Merged

Issue/4623 #4806

Merged

Build failed on master 1.41.0-nightly (abd69551b 2019-11-22) #4835

Closed

flip1995 added 3 commits November 22, 2019 13:47

Remove never_type feature

3c308b8

Stablized in rust-lang/rust#65355

Rustup to rust-lang/rust#66515

e4636f3

Fixes rust-lang#4835

Rustup to rust-lang/rust#66389

cadc35a

flip1995 force-pushed the rustup branch from 1b2f43d to cadc35a Compare November 22, 2019 12:47

flip1995 mentioned this pull request Nov 23, 2019

RIIR: Integration tests #4837

Merged

flip1995 added 2 commits November 23, 2019 16:47

Temporarily disable serde integration test

fb18354

ICEs because of rustc, not Clippy

Disable chalk integration test. Output too large

553db87

bors merged commit 553db87 into rust-lang:master Nov 23, 2019

bors mentioned this pull request Nov 23, 2019

Add custom ICE message that points to Clippy repo #4588

Merged

Manishearth deleted the rustup branch November 28, 2019 15:21

sinkuu mentioned this pull request Jan 8, 2020

Fix redundant_clone lint not working with PathBuf and OsString #5017

Merged

	let mir = cx.tcx.optimized_mir(def_id);

	if let Err((span, err)) = is_min_const_fn(cx.tcx, def_id, &mir) {

	fn visit_local(&mut self, local: &mir::Local, ctx: PlaceContext, _: mir::Location) {
	match ctx {
	PlaceContext::MutatingUse(MutatingUseContext::Drop) \| PlaceContext::NonUse(_) => return,
	_ => {},
	}

	if *local == self.local {
	self.used_other_than_drop = true;
	}
	}

	impl rustc_driver::Callbacks for ClippyCallbacks {
	fn config(&mut self, config: &mut interface::Config) {

Rustup to rustc 1.41.0-nightly (35ef33a8 2019-11-21) #4825

Rustup to rustc 1.41.0-nightly (35ef33a8 2019-11-21) #4825

Conversation

Manishearth commented Nov 18, 2019 • edited by flip1995 Loading

Manishearth commented Nov 18, 2019

This comment has been minimized.

flip1995 commented Nov 18, 2019 • edited Loading

Manishearth commented Nov 18, 2019

Manishearth commented Nov 18, 2019

flip1995 commented Nov 18, 2019

flip1995 commented Nov 18, 2019

Manishearth commented Nov 18, 2019

flip1995 commented Nov 18, 2019 • edited Loading

Manishearth commented Nov 18, 2019 via email

flip1995 commented Nov 18, 2019

Manishearth commented Nov 19, 2019

bors commented Nov 19, 2019

bors commented Nov 19, 2019

bors commented Nov 19, 2019

Manishearth commented Nov 19, 2019

flip1995 Nov 19, 2019 • edited Loading

Choose a reason for hiding this comment

flip1995 Nov 19, 2019

Choose a reason for hiding this comment

flip1995 Nov 19, 2019 • edited Loading

Choose a reason for hiding this comment

oli-obk Nov 19, 2019

Choose a reason for hiding this comment

This comment was marked as outdated.

flip1995 Nov 19, 2019

Choose a reason for hiding this comment

oli-obk Nov 19, 2019

Choose a reason for hiding this comment

flip1995 Nov 20, 2019

Choose a reason for hiding this comment

bors commented Nov 23, 2019

bors commented Nov 23, 2019

flip1995 commented Nov 23, 2019

bors commented Nov 23, 2019

bors commented Nov 23, 2019

flip1995 commented Nov 23, 2019

flip1995 commented Nov 23, 2019

flip1995 commented Nov 23, 2019

flip1995 commented Nov 23, 2019

bors commented Nov 23, 2019

bors commented Nov 23, 2019

bors commented Nov 23, 2019

Manishearth commented Nov 18, 2019 •

edited by flip1995

Loading

flip1995 commented Nov 18, 2019 •

edited

Loading

flip1995 commented Nov 18, 2019 •

edited

Loading

flip1995 Nov 19, 2019 •

edited

Loading

flip1995 Nov 19, 2019 •

edited

Loading