Fix ICE on undocumented_unsafe_blocks

[giraffate]

fix #7979

changelog: Fix ICE on undocumented_unsafe_blocks

[rust-highfive]

r? @camsteffen

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

[flip1995]

@bors r+

Thanks! And another one from this lint...

[bors]

📌 Commit 3f3d7c2 has been approved by flip1995

[bors]

⌛ Testing commit 3f3d7c2 with merge d550e5f...

[camsteffen]

This fixes the case in question but I think there are more cases unaccounted for. In general I think two spans should be ensured to have the same span.ctxt() before combining them with to, with_hi, etc, otherwise the operation is unpredictable. There can be any combination of contexts between the two spans.

A general fix might be...

if enclosing_scope_span.ctxt() != block_span.ctxt() {
    // get the span of the macro call
    // mac! { unsafe {} }
    enclosing_scope_span = hygiene::walk_chain(enclosing_scope_span, block_span.ctxt());
    if enclosing_scope_span.ctxt() != block_span.ctxt() {
        // if the contexts are still not equal after walk_chain,
        // the unsafe block span context is a descendant of the enclosing scope span context
        // { macro_with_unsafe!() }
        // return since I don't know what span to use
        return;
    }
}

But then I don't know if this lint is designed to work with the span of a macro call, where there can be an arbitrary amount of code between the start of the macro call and the unsafe block.

[camsteffen]

@flip1995 @giraffate

[bors]

☀️ Test successful - checks-action_dev_test, checks-action_remark_test, checks-action_test
Approved by: flip1995
Pushing d550e5f to master...