resolve: Partially convert ambiguous_glob_imports lint into a hard error

[petrochenkov]

I'm tired of this logic interfering with any attempts to fix or refactor glob imports.

The lint was implemented 2.3 years ago, and made deny-by-default and report-in-dependencies 4 months ago.

The removal is a bit over-eager because of one piece that wasn't implemented correctly (#113099 (comment)), but I want to look at the crater results.

Part of #114095.

[rustbot]

Some changes occurred in compiler/rustc_codegen_gcc

cc @antoyo, @GuillaumeGomez

[rustbot]

r? @nnethercote

rustbot has assigned @nnethercote.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[petrochenkov]

@bors try

[rust-bors]

☀️ Try build successful (CI)
Build commit: 446cb60 (446cb600aa4837dd6c513f14fa0d25a909b177d7, parent: e22dab387f6b4f6a87dfc54ac2f6013dddb41e68)

[petrochenkov]

@craterbot check

[craterbot]

👌 Experiment pr-149195 created and queued.
🤖 Automatically detected try build 446cb60
🔍 You can check out the queue and this experiment's details.

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[craterbot]

🚧 Experiment pr-149195 is now running

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[craterbot]

🎉 Experiment pr-149195 is completed!
📊 2676 regressed and 6 fixed (741888 total)
📊 2000 spurious results on the retry-regessed-list.txt, consider a retry¹ if this is a significant amount.
📰 Open the summary report.

⚠️ If you notice any spurious failure please add them to the denylist!
ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

Footnotes

1. re-run the experiment with crates=https://crater-reports.s3.amazonaws.com/pr-149195/retry-regressed-list.txt ↩

[nnethercote]

@petrochenkov: seems like a lot of regressions :( Should I review this?

[petrochenkov]

@petrochenkov: seems like a lot of regressions :( Should I review this?

Not yet.
@rustbot author

[rustbot]

Reminder, once the PR becomes ready for a review, use @rustbot ready.

[petrochenkov]

The largest part of the regressions are from openssl-0.10.*, it's a known issue that is addressed by #147984.
So this is blocked by #147984 now.
@rustbot blocked

[yaahc]

Correct me if I'm wrong on this one

@rustbot author

[petrochenkov]

@bors try

[rust-bors]

☀️ Try build successful (CI)
Build commit: aea72be (aea72be7e5474dab7f3e28cea3e2e6cfa3d1c3fa, parent: d0835adc4e114eac911150b3692b830b5583223a)

[petrochenkov]

@craterbot run mode=check-only p=1 crates=https://crater-reports.s3.amazonaws.com/pr-149195/retry-regressed-list.txt

[craterbot]

👌 Experiment pr-149195-3 created and queued.
🤖 Automatically detected try build aea72be
🔍 You can check out the queue and this experiment's details.

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[craterbot]

🚧 Experiment pr-149195-3 is now running

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[craterbot]

🎉 Experiment pr-149195-3 is completed!
📊 384 regressed and 0 fixed (4524 total)
📊 181 spurious results on the retry-regressed-list.txt, consider a retry¹ if this is a significant amount.
📰 Open the summary report.

⚠️ If you notice any spurious failure please add them to the denylist!
ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

Footnotes

1. re-run the experiment with crates=https://crater-reports.s3.amazonaws.com/pr-149195-3/retry-regressed-list.txt ↩

[petrochenkov]

Great, now all of the regressions are actually from turning ambiguous_glob_imports into a hard error, not from other effects.

There are no direct regressions, because ambiguous_glob_imports was already deny-by-default and apparently nobody allowed it.

There are regressions from dependencies though, with worst offenders being:

• noise-0.7.0 - 215 regressions out of 384 (noise-0.9.0 is the current version) @Razaekel
• rustybuzz-0.3.0 and rustybuzz-0.4.0 - 73 regressions out of 384 (rustybuzz-0.20.1 is the current version) @harfbuzz @behdad
• icentral-neighbors-0.1.0 - 38 regressions out of 384 (this is the current version) @klebs6
• pdf-0.8.1 and pdf-0.9.0 - 30 regressions out of 384 (this is the current version) @pdf-rs @s3bk

These four crates cover 356 regressions out of 384.
I guess I'll send fixes to icentral-neighbors and pdf, and just special-case identifiers from noise and rustybuzz to still produce a lint instead of an error.

[Razaekel]

@petrochenkov If I'm understanding this correctly, the issue in noise-0.7.0 is because of the struct in both Perlin and PerlinSurflet being named Perlin. Fixing this would require renaming one of the structs, which is what occurred in noise-0.8.0, because the rename was a breaking API change. As such, all the regression issues are due to crates continuing to use noise-0.7.0.

[behdad]

• rustybuzz-0.3.0 and rustybuzz-0.4.0 - 73 regressions out of 384 (rustybuzz-0.20.1 is the current version) @harfbuzz @behdad

I'm not sure what is being asked of me here, but rustybuzz is effectively deprecated and unmaintained. It's superseded by harfrust fork of it.

[petrochenkov]

@petrochenkov If I'm understanding this correctly, the issue in noise-0.7.0 is because of the struct in both Perlin and PerlinSurflet being named Perlin. Fixing this would require renaming one of the structs, which is what occurred in noise-0.8.0', because the rename was a breaking API change. As such, all the regression issues are due to crates continuing to use noise-0.7.0`.

@Razaekel
It should be possible to fix without an API breaking change, during the week I'll make a PR showing how.

[petrochenkov]

• rustybuzz-0.3.0 and rustybuzz-0.4.0 - 73 regressions out of 384 (rustybuzz-0.20.1 is the current version) @harfbuzz @behdad

I'm not sure what is being asked of me here, but rustybuzz is effectively deprecated and unmaintained. It's superseded by harfrust fork of it.

@behdad
Acknowledged, for now I'll just make an exception in the compiler for old versions of rustybuzz.

[behdad]

@behdad
Acknowledged, for now I'll just make an exception in the compiler for old versions of rustybuzz.

Thanks. Is there any chance to include https://github.com/harfbuzz/harfrust?

[petrochenkov]

@behdad
Acknowledged, for now I'll just make an exception in the compiler for old versions of rustybuzz.

Thanks. Is there any chance to include https://github.com/harfbuzz/harfrust?

I don't see any issues about this crate in the crater report, looks like it's not affected.

[behdad]

Thanks for checking.

[petrochenkov]

@petrochenkov If I'm understanding this correctly, the issue in noise-0.7.0 is because of the struct in both Perlin and PerlinSurflet being named Perlin. Fixing this would require renaming one of the structs, which is what occurred in noise-0.8.0', because the rename was a breaking API change. As such, all the regression issues are due to crates continuing to use noise-0.7.0`.

@Razaekel It should be possible to fix without an API breaking change, during the week I'll make a PR showing how.

Opened PR - Razaekel/noise-rs#369.

[petrochenkov]

• pdf-0.9.0 was fixed in fix future-incompatibility warning pdf-rs/pdf#259 and pdf-0.9.1 was published.
• icentral-neighbors-0.1.0 is fixed in Fix an ambigous glob import error klebs6/icentral-rs#1.

Some other crates (besides #149195 (comment)) causing multiple regressions:

• ark-test-curves-0.4.2 (7 regressions) - outdated version, ark-test-curves-0.5.0 doesn't have the issue
• polars-plan-0.27.2 (3 regressions) - outdated version, polars-plan-0.52.0 doesn't have the issue

[rustbot]

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.