Always emit unresolved import errors and hide unused import lint

[estebank]

Fix#63724.

r? @petrochenkov

[estebank]

@petrochenkov I originally had a similar intuition to what you mentioned, but in practice I found that we only have hard failures on the first time a module is not found. After that we always get Indeterminate, which makes 1) someself.finalize_import(indeterminate_import).is_some(), but not all, and 2) subsequent import errors are silenced. I don't fully understand why this is the case, as this behavior doesn't jive with what I would have expected the logic to do. The presented code is what I arrived at empirically to.

[petrochenkov]

(I need to make some experiments.)

[petrochenkov]

Mostly superseded by #64265.
I didn't include the "report all the indeterminate imports instead of one" part (which is also reasonable, but orthogonal), only the "unused" part.

[estebank]

Is the behavior of "report all the indeterminate imports as unresolved" correct? I can publish a PR after yours gets merged to do so. I think it is useful to have but not critical if it could lead to bad false positives (due to dependencies: use foo::bar; use bar::baz;).

[petrochenkov]

I don't think it's correct as implemented because finalize_import doesn't always return Some(err) for indeterminate imports, but they should be reported anyway.

What I think we should do is something roughly like this:

for import in determinate_imports.chain(indeterminate_imports){ifletSome(err) = finalize_import(import){report_error(err)}elseifis_indeterminate(import){report_error("cannot determine resolution")}}

[petrochenkov]

due to dependencies: use foo::bar; use bar::baz;

AFAIK, this case should be already covered by marking use foo::bar; as successfully resolved to Res::Err.

[bors]

☔ The latest upstream changes (presumably #64281) made this pull request unmergeable. Please resolve the merge conflicts.

[petrochenkov]

Thanks!
@bors r+

[bors]

📌 Commit 9a56187 has been approved by petrochenkov