Enforce short type string in more errors

[estebank]

error[E0529]: expected an array or slice, found `(..., ..., ..., ...)`
  --> $DIR/long-E0529.rs:10:9
   |
LL |     let [] = x;
   |         ^^ pattern cannot match with input type `(..., ..., ..., ...)`
   |
   = note: the full name for the type has been written to '$TEST_BUILD_DIR/$FILE.long-type-hash.txt'
   = note: consider using `--verbose` to print the full type name to the console

error[E0609]: no field `field` on type `(..., ..., ..., ...)`
  --> $DIR/long-E0609.rs:10:7
   |
LL |     x.field;
   |       ^^^^^ unknown field
   |
   = note: the full name for the type has been written to '$TEST_BUILD_DIR/$FILE.long-type-hash.txt'
   = note: consider using `--verbose` to print the full type name to the console

error[E0614]: type `(..., ..., ..., ...)` cannot be dereferenced
  --> $DIR/long-E0614.rs:10:5
   |
LL |     *x;
   |     ^^
   |
   = note: the full name for the type has been written to '$TEST_BUILD_DIR/$FILE.long-type-hash.txt'
   = note: consider using `--verbose` to print the full type name to the console

error[E0618]: expected function, found `(..., ..., ..., ...)`
  --> $DIR/long-E0618.rs:10:5
   |
LL | fn foo(x: D) {
   |        - `x` has type `(..., ..., ..., ...)`
LL |     x();
   |     ^--
   |     |
   |     call expression requires function
   |
   = note: the full name for the type has been written to '$TEST_BUILD_DIR/$FILE.long-type-hash.txt'
   = note: consider using `--verbose` to print the full type name to the console

Use multipart suggestion for cast -> into:

error[E0604]: only `u8` can be cast as `char`, not `u32`
  --> $DIR/E0604.rs:2:5
   |
LL |     1u32 as char;
   |     ^^^^^^^^^^^^ invalid cast
   |
help: try `char::from_u32` instead
   |
LL -     1u32 as char;
LL +     char::from_u32(1u32);
   |

CC #135919.

[rustbot]

r? @compiler-errors

rustbot has assigned @compiler-errors.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[compiler-errors]

r? compiler

[oli-obk]

It seems a bit awful to have to do this everywhere. Can't we do something similar to the other global settings that change formatting for a scope?

[estebank]

Yeah, I agree. My initial hope was to make it so only main messages needed to use this, but an even better approach would be to continue moving towards structured diagnostics and making the machinery handle the printing of Tys to always use short strings.

[estebank]

@oli-obk would you prefer we came up with a "hands off" solution that doesn't require as much redundant logic first, or land this while I work on that?

[oli-obk]

Yea, I think the added verbosity to diagnostics impls that will get reverted later is not worth it now and would def prefer going for another solution first

[estebank]

I'm encountering a small issue: TyCtxt contains the DiagCtxt, which gets constructed early. In order to properly shorten types, you need a FmtPrinter, which has a TyCtxt within it. I don't think that we can make it so DiagCtxt contains an optional TyCtxt (I'm assuming if I go down that route I'll end up with a cycle), which instead would require me to pass a tcx to every create_err and emit_err. There are >1500 of these. I created a new trait with two impls, one that shortens and one on () that does nothing, so when you don't have a TyCtxt it doesn't attempt to shorten, but I would still have to touch a significant part of the codebase to introduce this. And it would still only work for structured diagnostics. I think that for raw access I could introduce a macro so that instead of doing err.span_label(span, format!("{ty}")) we would do span_label!(tcx, err, span, "{ty}"), which should potentially allow me to customize the printing behavior when it comes to things that can be shortened, but whatever pattern I can think of, it will be intrusive one way or another :-/

[oli-obk]

Huh. The other schemes just use TLS I think?