@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) some self.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.

(I need to make some experiments.)

Mostly superseded by #64265.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.

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;).

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) {
    if let Some(err) = finalize_import(import) {
        report_error(err)
    } else if is_indeterminate(import) {
        report_error("cannot determine resolution")
    }
}

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.

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

Thanks!
@bors r+

📌 Commit 9a56187 has been approved by petrochenkov