Make ambiguous_glob_reexports an FCW at deny

[traviscross]

Is there any good reason for someone to intentionally create an ambiguous glob reexport? No such reasons come to mind offhand.

We lint strongly against use of these ambiguous glob reexports, so not linting strongly at the def-site pushes cost, relatively, to users rather than upstreams. It'd be better to push this cost upstream.

Let's make ambiguous_glob_reexports an FCW that lints at deny-by-default and that warns in dependencies.

TODO: I'm not fixing the tests at this point as I just want to do a crater run.

Related

• Lint ambiguous glob re-exports #107880
• Consider adding a lint for ambiguous re-exports, since they are unusable #107563
• Diesel stops building with nightly-2025-12-10 #149845
• Tracking Issue for ambiguous_glob_imports lint #114095
• resolve: Preserve ambiguous glob reexports in crate metadata #147984

[traviscross]

@bors try

[rust-bors]

💥 Test timed out after 21600s

[traviscross]

@bors try

[rust-bors]

☀️ Try build successful (CI)
Build commit: d336647 (d336647335db03a0fc5b24fc5325cd3905c5b8e8, parent: c9af9c1dc85a55e309345030ff9cb7ea247953fa)

[petrochenkov]

This is a lint and not a language rule because it uses the "effective visibility" tables (whether something is reachable from outside of the crate) for reporting, and those tables are not necessarily precise (can conservatively mark things as reachable).
All hard language rules currently use "local" nominal visibilities (what is written in pub), and not "global" effective visibilities (what is reachable where considering reexports, return types, etc).

So I'm skeptical about making this an FCW, but deny-by-default is probably fine.

[traviscross]

Makes sense.

@craterbot check

[craterbot]

👌 Experiment pr-151146 created and queued.
🤖 Automatically detected try build d336647
🔍 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