Position span label correctly when it isn't last

[estebank]

Fix #42595.

Before:

15 |     map.entry("e").or_insert(0) += 1;
   |     ---------------------------^^^^^ot use `+=` on type `&mut {integer}`

After:

15 |     map.entry("e").or_insert(0) += 1;
   |     ---------------------------^^^^^
   |     |
   |     cannot use `+=` on type `&mut {integer}`

[rust-highfive]

r? @eddyb

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

[Mark-Simulacrum]

I'm a little confused as to the intent of this message. Is it trying to just point at the += (which seems correct)? If so, the underline-like arrow beneath the rest of the line looks really odd to me. Might just not be used to it though.. nor do I have any ideas as to what else we could do.

[estebank]

@Mark-Simulacrum, the primary span which doesn't have a label is as follows:

15 |     map.entry("e").or_insert(0) += 1;
   |     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

My assumption is that there should be only one span in this error, either:

15 |     map.entry("e").or_insert(0) += 1;
   |     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ cannot use `+=` on type `&mut {integer}`

or

15 |     map.entry("e").or_insert(0) += 1;
   |     ^^^^^^^^^^^^^^^^^^^^^^^^^^^ cannot use `+=` on type `&mut {integer}`

Once we fix that, I'll have to find a new way to have a test for this bug, though.

[Mark-Simulacrum]

Oh. I would have expected the span to be on +=, actually, not the whole expression or some part of it. Obviously that's a separate discussion though -- no need to fix here.

[eddyb]

Argh why do I keep getting picked on these PRs. r? @jonathandturner

[sophiajt]

This is a Niko one, I think he did the drawing layout bit.

r? @nikomatsakis

[nikomatsakis]

it's progress for sure. I still find these sorts of overlap scenarios pretty mind boggling regardless, though i'm not sure how best to fix in general. I'd sort of like to separate the ^^^ and --- on different lines. i.e., if I was highlighting this by hand I would probably do:

a  bc  d
^^^^
    ----

but anyway.

[nikomatsakis]

I thought that "expected output" looked wrong...

[01:03:10] ---- test_snippet::multiple_labels_secondary_without_message_3 stdout ----

[01:03:10] 	span: Span { lo: BytePos(14), hi: BytePos(19), ctxt: #0 } label: "`a` is a good letter"

[01:03:10] text: Ok("a { b")

[01:03:10] span: Span { lo: BytePos(22), hi: BytePos(27), ctxt: #0 } label: ""

[01:03:10] text: Ok("c } d")

[01:03:10] expected output:

[01:03:10] ------

[01:03:10] error: foo

[01:03:10]  --> test.rs:3:3

[01:03:10]   |

[01:03:10] 3 |   a  bc  d

[01:03:10]   |   ^^^^----

[01:03:10]   |   |

[01:03:10]   |   `a` is a good letter

[01:03:10] 

[01:03:10] ------

[01:03:10] actual output:

[01:03:10] ------

[01:03:10] error: foo

[01:03:10]  --> test.rs:3:3

[01:03:10]   |

[01:03:10] 3 |   a { b { c } d }

[01:03:10]   |   ^^^^^   -----

[01:03:10]   |   |

[01:03:10]   |   `a` is a good letter

[01:03:10] 

[01:03:10] ------

r=me once this is fixed

[nikomatsakis]

@bors r+

[bors]

📌 Commit 8074a88 has been approved by nikomatsakis

[frewsxcv]

@bors rollup