Outer attribute after inner attribute error should give more information

[estebank]

Given the following

rust/src/test/ui/parser/inner-attr.rs

Lines 1 to 4 in dc6db14

	#[feature(lang_items)]
	#![recursion_limit="100"] //~ ERROR an inner attribute is not permitted following an outer attribute
	fn main() {}

we currently emit

rust/src/test/ui/parser/inner-attr.stderr

Lines 1 to 7 in dc6db14

	error: an inner attribute is not permitted following an outer attribute
	--> $DIR/inner-attr.rs:3:3
	|
	LL | #![recursion_limit="100"]
	| ^
	|
	= note: inner attributes, like `#![no_std]`, annotate the item enclosing them, and are usually found at the beginning of source files. Outer attributes, like `#[test]`, annotate the item following them.

We should emit at least the following

error: an inner attribute is not permitted following an outer attribute
  --> $DIR/inner-attr.rs:3:3
   |
LL | #[feature(lang_items)]
   | ---------------------- previous outer attribute
LL | 
LL | #![recursion_limit="100"]
   | ^^^^^^^^^^^^^^^^^^^^^^^^^ not permited following an outer attibute
   |
   = note: inner attributes, like `#![no_std]`, annotate the item enclosing them, and are usually found at the beginning of source files. Outer attributes, like `#[test]`, annotate the item following them.

We might also want to provide suggestions when it makes sense (like turning the feature attr into an outer attr).

[robinboehm]

Hey @estebank ,

I tried to solve this with a little help of @badboy on a rust meetup.
I wasn't able to run cargo test because of the following message.

rust git:(master) ✗ RUST_BACKTRACE=1 cargo test src/test/ui/parser
   Compiling rustc_codegen_llvm v0.0.0 (/Users/robinboehm/development/rust/rust/src/librustc_codegen_llvm)
   Compiling libc v0.2.54
   Compiling compiler_builtins v0.1.15
   Compiling profiler_builtins v0.0.0 (/Users/robinboehm/development/rust/rust/src/libprofiler_builtins)
   Compiling strip-ansi-escapes v0.1.0
error: failed to run custom build command for `profiler_builtins v0.0.0 (/Users/robinboehm/development/rust/rust/src/libprofiler_builtins)`

Caused by:
  process didn't exit successfully: `/Users/robinboehm/development/rust/rust/target/debug/build/profiler_builtins-48a5ce79ce969ba5/build-script-build` (exit code: 101)
--- stderr
thread 'main' panicked at 'called `Option::unwrap()` on a `None` value', src/libcore/option.rs:347:21
stack backtrace:
   0: <unknown>
   1: <unknown>
   2: <unknown>
   3: <unknown>
   4: <unknown>
   5: <unknown>
   6: <unknown>
   7: <unknown>
   8: <unknown>
   9: <unknown>
  10: <unknown>
  11: <unknown>
  12: <unknown>
  13: <unknown>
  14: <unknown>

warning: build failed, waiting for other jobs to finish...
error: build failed

But I started to add some code you can read here: robinboehm@d71319c

is this going into the right direction?

[estebank]

@robinboehm thank you for taking a look at this!

You should be able to run the tests with ./x.py test src/test/ui --stage 1. Use ./x.py test src/test/ui --stage 1 --bless to update the relevant .stderr files.

---

The code you've written would generate the following output:

note: previous outer attribute
   |
LL | #![recursion_limit="100"]
   | ^

error: an inner attribute is not permitted following an outer attribute
  --> $DIR/inner-attr.rs:3:3
   |
LL | #![recursion_limit="100"]
   |  ^
   |
   = note: inner attributes, like `#![no_std]`, annotate the item enclosing them, and are usually found at the beginning of source files. Outer attributes, like `#[test]`, annotate the item following them.

Instead, we'd want this to be a single error. What you'd want to do is 1) get the span for the prior attribute from attrs.last() 2) expand InnerAttributeParsePolicy::NotPermitted to also hold an extra field (Span, &str) and 3) change the following

rust/src/libsyntax/parse/attr.rs

Lines 94 to 104 in dc6db14

	if let InnerAttributeParsePolicy::NotPermitted { reason } = inner_parse_policy
	{
	let span = self.span;
	self.diagnostic()
	.struct_span_err(span, reason)
	.note("inner attributes, like `#![no_std]`, annotate the item \
	enclosing them, and are usually found at the beginning of \
	source files. Outer attributes, like `#[test]`, annotate the \
	item following them.")
	.emit()
	}

to be

                        self.diagnostic()
                            .struct_span_err(span, reason)
                            .span_label(prev_attr_sp, prev_attr_msg)
                            .note("inner attributes, like `#![no_std]`, annotate the item \
                                   enclosing them, and are usually found at the beginning of \
                                   source files. Outer attributes, like `#[test]`, annotate the \
                                   item following them.")
                            .emit()

This lets you point at the prior span in the same error.

The second part of this ticket is changing the linked code to no longer use self.prev_span, but rather synthesize a new span from the first #'s span to the closing ]. You can use beginning_span.to(end_span) to do so. Finally, we change the above to be something like

                        self.diagnostic()
                            .struct_span_err(span, reason)
                            .span_label(span, reason)
                            .span_label(prev_attr_sp, prev_attr_msg)
                            .note("inner attributes, like `#![no_std]`, annotate the item \
                                   enclosing them, and are usually found at the beginning of \
                                   source files. Outer attributes, like `#[test]`, annotate the \
                                   item following them.")
                            .emit()```

[YairHalberstadt]

@robinboehm

Are you still working on this?

I'm thinking of giving it a try.

[robinboehm]

Go on! I don’t have so much time atm so just go there step by step while learning rust. If you want so solve it - awesome 🙌

…

On Thu 6. Jun 2019 at 09:51, Yair Halberstadt ***@***.***> wrote: @robinboehm <https://github.com/robinboehm> Are you still working on this? I'm thinking of giving it a try. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#61218?email_source=notifications&email_token=AAGIJJ5XRIILBBLLXLPIQBTPZC6X5A5CNFSM4HPXMLD2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODXCA5QY#issuecomment-499388099>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAGIJJ72VC2S4Y3P2I7RUJDPZC6X5ANCNFSM4HPXMLDQ> .

-- *Angular.DE - Schulungen & Beratung* *Veranstalter der NG-DE 2019* Ticket für Angular-Konferenz NG-DE im August in Berlin kaufen: https://angular.de/ng-de-konferenz/ <https://angular.de/ng-de-konferenz/> Angular-Schulung für Einsteiger buchen: https://angular.de/schulungen/angular-intensiv/ <https://angular.de/schulungen/angular-intensiv/> Angular-Schulung für Fortgeschrittene buchen: https://angular.de/schulungen/angular-advanced/ <https://angular.de/schulungen/angular-advanced/> Symetics GmbH, c/o Factory, Lohmühlenstraße 65, 12435 Berlin Amtsgericht Essen, HRB 26120, Sitz Essen Geschäftsführer: Sascha Brink, Robin Böhm