Point to enclosing block/fn on nested unsafe

[estebank]

When declaring nested unsafe blocks (unsafe {unsafe {}}) that trigger
the "unnecessary unsafe block" error, point out the enclosing unsafe block or unsafe fn that makes it unnecessary.

Fixes #39144.

[rust-highfive]

r? @arielb1

(rust_highfive has picked a reviewer for you, use r? to override)

[estebank]

Another option is to use span_label instead of span_help:

[durka]

Might want to carve out an exception for unsafe blocks within macros.

[estebank]

cc @nikomatsakis @jonathandturner

[nikomatsakis]

@estebank this seems nice; but it'd be nicer still if we used a more targeted span (i.e., just the unsafe keyword).

[estebank]

@nikomatsakis There's only one reason I'm reticent to that suggestion: The current presentation will only be as shown if the block is less than 8 lines (showing only the first line that contains the keyword otherwise), and I'm looking into making the multiline presentation smarter about longer blocks (instead of using the first line, showing the first X and last Y lines). I feel it is useful to know both the start and end of the enclosing block (as not everyone will know about "jump to closing brace" in their editor) when possible.

What do you think?

[nikomatsakis]

@estebank I think people can find the end of their blocks ok, personally. But also, I sort of like the "collapsed" error where both of the unsafe blocks are shown simultaneously, but

• we should make sure to put a label on the primary span (i.e., ^^^ unnecessary unsafe block), since it looks really weird otherwise
• I guess it just seems kind of heavy with the ascii art; I'm very happy we have it, but I'd prefer for it to be more unusual

[arielb1]

Personally, I don't like errors with separate spans - it blasts the console too much.

[arielb1]

The code seems ok. I think r? @jonathandturner

[sophiajt]

Looks good.

@bors r+

[bors]

📌 Commit f31c8b7 has been approved by jonathandturner

[bors]

⌛ Testing commit f31c8b7 with merge 2ae8de7...

[bors]

💔 Test failed - status-travis

[sophiajt]

Seems like a legit failure:

error: no field `map` on type `rustc::ty::TyCtxt<'_, '_, '_>`

   --> /checkout/src/librustc_lint/unused.rs:194:36

    |

194 |             let parent_id = cx.tcx.map.get_parent_node(id);

    |                                    ^^^ unknown field

error: no field `map` on type `rustc::ty::TyCtxt<'_, '_, '_>`

   --> /checkout/src/librustc_lint/unused.rs:201:30

    |

201 |                 })) = cx.tcx.map.find(parent_id) {

    |                              ^^^ unknown field

error: no field `map` on type `rustc::ty::TyCtxt<'_, '_, '_>`

   --> /checkout/src/librustc_lint/unused.rs:219:41

    |

219 |                     db.span_note(cx.tcx.map.span(id),

    |                                         ^^^ unknown field

[hanna-kruppe]

FYI: The fix for this is simply using tcx.hir instead of tcx.map.

[estebank]

@jonathandturner fixed merge conflict and added label to the block.

[sophiajt]

@bors r+

[bors]

📌 Commit 3a9560a has been approved by jonathandturner

[bors]

🔒 Merge conflict

[alexcrichton]

@bors: retry

[bors]

🔒 Merge conflict

[alexcrichton]

@bors: r=jonathandturner

[bors]

📌 Commit ac2bc7c has been approved by jonathandturner