Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

try reading rust-version from Cargo.toml #8774

Merged
merged 4 commits into from
Jun 28, 2022

Conversation

hellow554
Copy link
Contributor

@hellow554 hellow554 commented May 2, 2022

Cargo.toml can contain a field rust-version, that acts like a MSRV of
clippy.toml file: https://doc.rust-lang.org/cargo/reference/manifest.html#the-rust-version-field
This will try to read that field and use it, if the clippy.toml config
has no msrv entry

changelog: respect rust-version from Cargo.toml

closes #8746
closes #7765

@rust-highfive
Copy link

r? @giraffate

(rust-highfive has picked a reviewer for you, use r? to override)

@rust-highfive rust-highfive added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties label May 2, 2022
@flip1995
Copy link
Member

flip1995 commented May 2, 2022

This sounds familiar. I think I also worked on that before 🤔

....

I did: rust-lang/cargo#9967 and I claimed #7765 back then. I probably forgot about finishing it 😄

I think one issue I couldn't resolve with this is that MetadataCommand is a quite expensive operation. This isn't really noticeable if compiling a single crate, but for example when running Clippy on rustc test suite, this can increase the time it takes immensely. See #5505 (comment)

cc @matthiaskrgr Do you still have a way to run Clippy on rustc's test suite to test this branch and whether it brings back the issue linked above?

@hellow554
Copy link
Contributor Author

I really tried searching for a duplicate, I'm sorry :D
But thanks for bringing it up my attention.
Should I add a closes #7765 as well?

@flip1995
Copy link
Member

flip1995 commented May 2, 2022

Should I add a closes #7765 as well?

Yes, please.

Copy link
Member

@flip1995 flip1995 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm also not sure about the precedence. Should the rust-version in the Cargo.toml file be valued more or the one in the clippy.toml file? Should we deprecate the msrv conf option and point to the rust-version instead?

Comment on lines 481 to 482
Err(e) => {
sess.struct_err(&format!("could not read cargo metadata: {}", e)).emit();
Copy link
Member

@flip1995 flip1995 May 2, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think this is a hard error. For example when using clippy-driver outside of a cargo project, no Cargo.toml file can be found. I would just silently keep conf.msrv = None here.

@hellow554
Copy link
Contributor Author

I'm also not sure about the precedence. Should the rust-version in the Cargo.toml file be valued more or the one in the clippy.toml file? Should we deprecate the msrv conf option and point to the rust-version instead?

Good question. I thought, that the clippy.toml file should have a higher precedence, than the cargo.toml file. Deprecating it would be another idea, yes... But that's above my paygrade 😇

@flip1995
Copy link
Member

flip1995 commented May 2, 2022

Let's talk about this in tomorrows meeting 👍

@flip1995 flip1995 added the I-nominated Issue: Nominated to be discussed at the next Clippy meeting label May 2, 2022
@ghost
Copy link

ghost commented May 3, 2022

I think one issue I couldn't resolve with this is that MetadataCommand is a quite expensive operation. This isn't really noticeable if compiling a single crate, but for example when running Clippy on rustc test suite, this can increase the time it takes immensely. See #5505 (comment)

Shouldn't Cargo provide rust-version as an environment variable? They already provide a lot of other manifest data in that way - cargo docs.

We already use CARGO_PRIMARY_PACKAGE in the driver.

let in_primary_package = env::var("CARGO_PRIMARY_PACKAGE").is_ok();

@flip1995
Copy link
Member

flip1995 commented May 3, 2022

We discussed this in today's meeting. We came to the agreement that:

  • We shouldn't deprecate Clippy's MSRV config option (yet)
  • The clippy.toml MSRV config option should have precedence.
  • If both clippy.toml MSRV and Cargo.toml rust-version are defined, but differ, a warning should be emitted.

I think the current implementation covers the first two points. Do you mind adding the warning? If you need help with that let me know.

Also we should answer the perf question and if cargo providing this through a env var is and option. Maybe we can even get that from rustc itself?

@hellow554
Copy link
Contributor Author

hellow554 commented May 4, 2022

Thanks for the discussion on this topic.

Currently, I only use the expensive MetadataCommand call if clippy.toml hasn't that msrv option set. If I want to check them both, I need to make that call every time. I can do that, but that will very much likely make things worse.

I can open an issue/pr @ https://github.com/rust-lang/cargo to insert a CARGO_PKG_RUST_VERSION field to the environment.
I'm not sure what you mean by "getting it from rustc itself"? rustc doesn't read a Cargo.toml IIRC, only cargo does.

@flip1995
Copy link
Member

flip1995 commented May 4, 2022

If I want to check them both, I need to make that call every time. I can do that, but that will very much likely make things worse.

Not really, because the issue only occurs if you check multiple single files, not a cargo project. But if you don't have a Cargo.toml for those single files you also won't have a clippy.toml. Maybe it's a none-issue. Let me check that.

'm not sure what you mean by "getting it from rustc itself"? rustc doesn't read a Cargo.toml IIRC, only cargo does.

For some reason I had in mind that a #[rust-version] attribute exists and that therefore rustc must have some information about that. But that's not the case, so nevermind.

@flip1995
Copy link
Member

flip1995 commented May 4, 2022

Checking this with flamegraph, the MetadataCommand takes up 3-4% of the runtime, when running on a single file.

Looking at cargo uitest this has some more impact:

Without this patch:

cargo uitest > /dev/null  75.74s user 26.33s system 886% cpu 11.511 total
cargo uitest > /dev/null  75.96s user 26.91s system 891% cpu 11.545 total
cargo uitest > /dev/null  76.83s user 26.69s system 879% cpu 11.768 total

With this patch:

cargo uitest > /dev/null  90.80s user 30.94s system 883% cpu 13.785 total
cargo uitest > /dev/null  91.13s user 30.94s system 880% cpu 13.864 total
cargo uitest > /dev/null  91.20s user 31.08s system 892% cpu 13.697 total

(meassured on Intel Core i7-9750H @ 12x 4.5GHz)

It isn't too bad, but definitely a noticable regression.

I think creating an issue in cargo about the env var or (maybe better) asking on Zulip about it would be good. I'm open to merge this with MetadataCommand if the cargo team doesn't want to add the env var.

@hellow554
Copy link
Contributor Author

Thanks for the measurement!
I'm not really firm with zulip, would you mind asking there? I would try to prepare a PR for that?

@flip1995
Copy link
Member

flip1995 commented May 4, 2022

If no such env var should get added we could just check if CARGO_MANIFEST_DIR is set before running MetadataCommand. That way we can be sure, that Clippy is run through cargo and not through clippy-driver. But that's only a workaround.

@hellow554
Copy link
Contributor Author

So I did things™.

Instead of using CargoMetadata I use what @bjorn3 suggested and reading the $CARGO_MANIFEST_DIR/Cargo.toml file directly as toml file and parse the package.rust-version field from it.
This should give a little speed advantage over the cargo metadata call.

What I wanted next, but couldn't achieve because of the way the Conf struct is created, is to have the RustcVersion directly, instead of storing a string and parsing it all over again. But for this to work I either need to implement Deserialize directly for that struct (looking @flip1995 ;) ) or finding a clever workaround...

@hellow554
Copy link
Contributor Author

I really would like to add some tests, but I'm not really sure how to integrate them into the clippy testsuite. After all, I would like to have at least 7 different testcases and therefore 7 different cargo projects.

file-attr clippy cargo
None None None
None None Some
None Some None
...

Copy link
Member

@flip1995 flip1995 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

But for this to work I either need to implement Deserialize directly for that struct (looking @flip1995 ;) ) or finding a clever workaround

I intentionally wrote the rustc-version crate without deps. So adding serde for this isn't really an option.

You can add tests for this in the tests/ui-toml or tests/ui-cargo dirs. See for example: https://github.com/rust-lang/rust-clippy/tree/master/tests/ui-cargo/cargo_common_metadata/pass which has both, a Cargo.toml and a clippy.toml file in the test.

clippy_lints/src/lib.rs Outdated Show resolved Hide resolved
clippy_lints/src/lib.rs Outdated Show resolved Hide resolved
@flip1995
Copy link
Member

flip1995 commented May 5, 2022

Instead of using CargoMetadata I use what @bjorn3 suggested and reading the $CARGO_MANIFEST_DIR/Cargo.toml file directly as toml file and parse the package.rust-version field from it.

Not sure if we should re-parse it by hand or just check for the CARGO_MANIFEST_DIR env var and then call MetadataCommand.

@hellow554
Copy link
Contributor Author

Not sure if we should re-parse it by hand or just check for the CARGO_MANIFEST_DIR env var and then call MetadataCommand.

I mean, in the end, the two options should be equal in their outcome.
Where they can differ is the speed. As said, I avoided using MetadataCommand, because of the speed concerns. Parsing a toml file should be relativly easy and in the end, I'm doing the same as cargo does, am I right? :/

@flip1995
Copy link
Member

flip1995 commented May 5, 2022

Where they can differ is the speed. As said, I avoided using MetadataCommand, because of the speed concerns. Parsing a toml file should be relativly easy and in the end, I'm doing the same as cargo does, am I right? :/

Yeah I guess this is only style preference in the end. I don't think the perf difference matters much when linting a whole crate with cargo clippy. The perf concern is only with clippy-driver single-file.rs.

Let's keep the toml parsing code for now 👍

@matthiaskrgr
Copy link
Member

@flip1995 sorry, just randomly saw the mention (I'm getting like 30 github mails a day and don't check them all tbh 😓 )
I think the iterations were tacked in here: #3142
(from bash for-loop to a semi-fledged rust program (icemaker))
icemaker has a --clippy mode but the setup might be a bit tricky (you'll need RTIM master toolchain + clippy installed with the name "master" and systemd-run at least)

@hellow554
Copy link
Contributor Author

I did the suggested changes and rebased everything into one commit.

Hope this is ready for merging :)

Copy link
Member

@flip1995 flip1995 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM. Do you want to add some tests to tests/ui-cargo?

clippy_lints/src/lib.rs Outdated Show resolved Hide resolved
@hellow554
Copy link
Contributor Author

Hey @flip1995 I need your help again.

I seem to be stuck to the same problem as last time, when I tried to parse Cargo.toml manually, namely that when running tests I don't get the rust-version from the tests Cargo.toml, but from the project one.

Running pass_both_same yields an empty string, but when I change the Cargo.toml file in the clippy root directory to something like 1.62.51 I get exactly that, but not the one from the tests Cargo.toml file.

Is there a way around this? Else I would have to use cargo-metadata again and the whole point failed :(
Is that a problem in cargo uitest?

@flip1995
Copy link
Member

We may want to look at the UI-tests and if we can change something there to test it. We should definitely use the env var and not cargo_metadata. I'll have to look more in depth into that though...

@flip1995 flip1995 added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties and removed S-blocked Status: marked as blocked ❌ on something else such as an RFC or other implementation work labels Jun 18, 2022
@ghost
Copy link

ghost commented Jun 19, 2022

Would making a UI test with compiletest header to set the environment variable work?

@hellow554
Copy link
Contributor Author

I think I finally found the solution. It passes on my machine™.

Problem is, that compiletest_rs does not parse the Cargo.toml file in the tests folder itself.
I see two options: either use the compiletest-header as @mikerite suggests, or, to stick with the current theme, we just parse the Cargo.toml file ourself with the toml crate and set the environment variable ourself.

I know, this isn't a perfect solution, but this is the best I can do :)

Comment on lines -1 to +2
Using config file `$SRC_DIR/tests/ui-cargo/multiple_config_files/warn/.clippy.toml`
Warning: `$SRC_DIR/tests/ui-cargo/multiple_config_files/warn/clippy.toml` will be ignored.
Using config file `$SRC_DIR/.clippy.toml`
Warning: `$SRC_DIR/clippy.toml` will be ignored.
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is a "fallout" by setting CARGO_MANIFEST_DIR, but I think this is the proper way

Copy link
Member

@flip1995 flip1995 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Impl LGTM and also the solution to the tests. Only some minor things left.

Cargo.toml Outdated Show resolved Hide resolved
README.md Show resolved Hide resolved
clippy_lints/src/lib.rs Outdated Show resolved Hide resolved
@@ -0,0 +1,13 @@
#![feature(custom_inner_attributes)]
#![clippy::msrv = "1.58.0"]
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

So, this doesn't produce a warning. That is not optimal, but I wouldn't try to address this in this PR.

Can you add a FIXME, that this should produce some kind of warning, that it overrides the value from one of the toml files?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maybe abandon that attribute in the long run? ;)

@hellow554
Copy link
Contributor Author

Done

@hellow554 hellow554 force-pushed the cargo-rust-version branch 3 times, most recently from 4883039 to f0a1cd5 Compare June 28, 2022 06:01
Cargo.toml can contain a field `rust-version`, that acts like a MSRV of
clippy.toml file: https://doc.rust-lang.org/cargo/reference/manifest.html#the-rust-version-field
This will try to read that field and use it, if the clippy.toml config
has no `msrv` entry
compiletest_rs is not meant to test full cargo projects, but instead
only files.
So we need to parse the `Cargo.toml` file ourself and set the
corresponding environment variable. In this case we just set
`CARGO_PKG_RUST_VERSION`, nothing more. But, of course, this can be
extended.
@hellow554
Copy link
Contributor Author

CI is green and I think this is (finally) ready to merge

@flip1995
Copy link
Member

Thanks so much for this and your patience with all the back and forth!

@bors r+

@bors
Copy link
Collaborator

bors commented Jun 28, 2022

📌 Commit f0a1cd5 has been approved by flip1995

@bors
Copy link
Collaborator

bors commented Jun 28, 2022

⌛ Testing commit f0a1cd5 with merge b776fb8...

@bors
Copy link
Collaborator

bors commented Jun 28, 2022

☀️ Test successful - checks-action_dev_test, checks-action_remark_test, checks-action_test
Approved by: flip1995
Pushing b776fb8 to master...

@bors bors merged commit b776fb8 into rust-lang:master Jun 28, 2022
jplatte added a commit to ruma/ruma that referenced this pull request Jul 19, 2022
Clippy now respects the rust-version field in Cargo manifests:
rust-lang/rust-clippy#8774
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
S-waiting-on-review Status: Awaiting review from the assignee but also interested parties
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Suggestion: use rust-version field Support the rust-version field in Cargo.toml to specify the MSRV
6 participants