annotate-snippet emitter: Add UI test where annotated_files count is > 1

[phansch]

Part of #59346

It would be nice to have a UI test for the annotate-snippet emitter, that includes annotations that span multiple files. The goal of this issue is to find and add such a test.

Instructions

A simple way to look for this, would be to add this code:

if annotated_files.len() > 1 {
    let filename = primary_lo.file.name.to_string();
    eprintln!("annotated_files > 1 in {}", filename);
}

to the top of fn slices_for_files and then check the failing UI tests.

If there are failing UI tests, it means that those UI tests include a diagnostic that probably spans multiple files.

The next step, is to copy that file (and possible auxiliary files) over into src/test/ui/annotate-snippet/ and add // compile-flags: --error-format human-annotate-rs to the top of the UI test.

Once that's done, be sure to run the added test and open a PR including the blessed stderr files.

[phansch]

Under the assumption, that we do have an existing UI test with annotations from multiple files, I'm happy to help out anyone who wants to take on this issue. Otherwise we'll have to write a test case ourselves.

@rustbot modify labels to +A-diagnostics, +E-mentor, +E-help-wanted

[bIgBV]

Hey @phansch I'd like to pick this issue up.

[phansch]

@bIgBV sure, it's all yours! Let me know if you have any questions. The rustc-guide should help you to get started working on the compiler.

[bIgBV]

@phansch I made the change suggested in the issue and ran all the ui tests (over 9000!). None of them failed, meaning a new test needs to be written for this specific usecase, correct?

[phansch]

Thanks! Unfortunately I just realized I made a mistake in the instructions :/
If it's included in annotate_snippet_emitter_writer.rs it will only get executed for UI tests that use the new work-in-progress emitter, which only has one UI test currently

Instead, you'll want to include the snippet below this line in emitter.rs:

rust/src/librustc_errors/emitter.rs

Line 1172 in 7eb65df

let mut annotated_files = FileWithAnnotatedLines::collect_annotations(msp, &self.sm);

If you could give that a try too, that would be great!

[bIgBV]

Will do!

EDIT: @phansch re-ran it with the new changes, still no failures. I'm guessing that we need to add a new test.

[bIgBV]

@phansch looks like edits on github don't generate notifications, apologies. But I retried the tests with the second change you suggested. Still no failing tests.

[phansch]

@bIgBV Sorry, been out sick the past couple of days. Thanks for testing again!

It's weird that you don't get failing tests 🤔 Could you try to apply this specific patch?

diff --git a/src/librustc_errors/emitter.rs b/src/librustc_errors/emitter.rs
index 0ce69eecc6..8a3fbb6e09 100644
--- a/src/librustc_errors/emitter.rs
+++ b/src/librustc_errors/emitter.rs
@@ -1185,6 +1185,10 @@ impl EmitterWriter {
             emit_to_destination(&buffer.render(), level, &mut self.dst, self.short_message)?;
             return Ok(());
         };
+        if annotated_files.len() > 1 {
+            let filename = primary_lo.file.name.to_string();
+            panic!("annotated_files > 1 in {}", filename);
+        }
         if let Ok(pos) =
             annotated_files.binary_search_by(|x| x.file.name.cmp(&primary_lo.file.name)) {
             annotated_files.swap(0, pos);

When you run ./x.py test src/test/ui with the patch you should see at least 5 failing UI tests (At least I do, but I didn't have time to take a closer look, yet)

[bIgBV]

Okay, something is weird, after pulling in the latest master and making the exact changes in the patch you provided, there are over 8000 failing tests. I want to check if there is something obviously wrong going on, but I might not get back until tomorrow, as each test run takes close to an hour on my laptop.

[bIgBV]

@phansch sorry for the late reply, I got a little caught up but finally got around to making the changes and doing a fresh build and test. there are 18 failing tests on the latest stable rustc build.

failures:
    [ui] ui/async-await/issues/issue-62009-1.rs
    [ui] ui/closures/closure-move-sync.rs
    [ui] ui/cross/cross-file-errors/main.rs
    [ui] ui/derives/derives-span-Hash-enum-struct-variant.rs
    [ui] ui/derives/derives-span-Hash-enum.rs
    [ui] ui/derives/derives-span-Hash-struct.rs
    [ui] ui/derives/derives-span-Hash-tuple-struct.rs
    [ui] ui/editions/edition-keywords-2018-2015-parsing.rs
    [ui] ui/editions/edition-keywords-2018-2018-parsing.rs
    [ui] ui/impl-trait/impl-generic-mismatch.rs
    [ui] ui/interior-mutability/interior-mutability.rs
    [ui] ui/issues/issue-21160.rs
    [ui] ui/macro_backtrace/main.rs
    [ui] ui/no-send-res-ports.rs
    [ui] ui/proc-macro/multispan.rs
    [ui] ui/proc-macro/three-equals.rs
    [ui] ui/rfc-1937-termination-trait/termination-trait-test-wrong-type.rs
    [ui] ui/traits/trait-suggest-where-clause.rs

test result: FAILED. 8993 passed; 18 failed; 35 ignored; 0 measured; 0 filtered out

I'm guessing I need to move on to the next step as you detailed in the issue right? Could you give me some context into what an annotated snippet is and why it matters that it spans multiple files?

[phansch]

sorry for the late reply

Same here 😅

I'm guessing I need to move on to the next step as you detailed in the issue right?

Yup, at least one of those tests should fail with our panic message. I think
src/test/ui/proc-macro/multispan.rs would be a good candidate to copy.
Apart from the instructions, I think you'll also have to make sure to update the
path to the auxiliary file in the test header or copy the auxiliary file, too.
(The aux header allows to specify a second (or more) local crate to be compiled
so that it can be used in the test)

What is an annotated snippet?

In general, a snippet is a piece of code that can span multiple lines and
potentially files. An annotation points at a portion of the
snippet and contains annotations for parts of the snippet.

Aside: Note that rustc has a Line type instead of a Snippet type. The new
diagnostics library has a Snippet type instead of a line type.
That's why we have to convert from one to the other for #59346.

We care about the fact that a snippet can span more than one file, because the
new diagnostics emitting library might output those cases differently. This
issue is essentially about finding out how that output looks.

Let me know if you have more questions or need help (You can also ask on Zulip in #t-compiler/wg-diagnostics)

[hdhoang]

ping from triage @bIgBV, any updates on this? Thanks for your work!

[bIgBV]

@hdhoang sorry about the delay, I got a bit busy with work recently, but I'm going to pick this up again.

[Enselic]

PR with fix: #114018