no errors encountered even though `delay_span_bug` issued

[WaffleLapkin]

Reproducible at teloxide-core#828ab071.

; cargo +nightly-2022-01-24 fmt
thread 'main' panicked at 'no errors encountered even though `delay_span_bug` issued', compiler/rustc_errors/src/lib.rs:1188:13
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

Backtrace

; cargo +nightly-2022-01-24 fmt
thread 'main' panicked at 'no errors encountered even though `delay_span_bug` issued', compiler/rustc_errors/src/lib.rs:1188:13
stack backtrace:
   0: rust_begin_unwind
             at /rustc/84322efad553c7a79c80189f2d1b9197c1aa005f/library/std/src/panicking.rs:577:5
   1: core::panicking::panic_fmt
             at /rustc/84322efad553c7a79c80189f2d1b9197c1aa005f/library/core/src/panicking.rs:135:14
   2: core::panicking::panic_display::<&str>
   3: <rustc_errors::HandlerInner>::flush_delayed
   4: <rustc_errors::HandlerInner as core::ops::drop::Drop>::drop
   5: <rustfmt_nightly::parse::session::ParseSess>::set_silent_emitter
   6: rustfmt_nightly::formatting::format_project::<rustfmt_nightly::Session<alloc::vec::Vec<u8>>>
   7: <scoped_tls::ScopedKey<rustc_span::SessionGlobals>>::with::<<rustfmt_nightly::Session<alloc::vec::Vec<u8>>>::format_input_inner::{closure#0}, core::result::Result<rustfmt_nightly::FormatReport, rustfmt_nightly::ErrorKind>>
   8: <rustfmt_nightly::Session<alloc::vec::Vec<u8>>>::format_input_inner
   9: std::panicking::try::<core::option::Option<rustfmt_nightly::FormattedSnippet>, rustfmt_nightly::format_snippet::{closure#0}>
  10: rustfmt_nightly::format_snippet
  11: rustfmt_nightly::format_code_block
  12: <rustfmt_nightly::macros::MacroBranch>::rewrite
  13: <rustfmt_nightly::lists::ListItems<core::slice::iter::Iter<rustfmt_nightly::macros::MacroBranch>, rustfmt_nightly::macros::rewrite_macro_def::{closure#1}, rustfmt_nightly::macros::rewrite_macro_def::{closure#2}, rustfmt_nightly::macros::rewrite_macro_def::{closure#3}> as core::iter::traits::iterator::Iterator>::next
  14: <alloc::vec::Vec<rustfmt_nightly::lists::ListItem> as alloc::vec::spec_from_iter::SpecFromIter<rustfmt_nightly::lists::ListItem, rustfmt_nightly::lists::ListItems<core::slice::iter::Iter<rustfmt_nightly::macros::MacroBranch>, rustfmt_nightly::macros::rewrite_macro_def::{closure#1}, rustfmt_nightly::macros::rewrite_macro_def::{closure#2}, rustfmt_nightly::macros::rewrite_macro_def::{closure#3}>>>::from_iter
  15: rustfmt_nightly::macros::rewrite_macro_def
  16: <rustfmt_nightly::visitor::FmtVisitor>::visit_item
  17: <rustfmt_nightly::visitor::FmtVisitor>::visit_items_with_reordering
  18: <rustfmt_nightly::visitor::FmtVisitor>::format_separate_mod
  19: rustfmt_nightly::formatting::format_project::<rustfmt_nightly::Session<std::io::stdio::Stdout>>
  20: <scoped_tls::ScopedKey<rustc_span::SessionGlobals>>::with::<<rustfmt_nightly::Session<std::io::stdio::Stdout>>::format_input_inner::{closure#0}, core::result::Result<rustfmt_nightly::FormatReport, rustfmt_nightly::ErrorKind>>
  21: <scoped_tls::ScopedKey<rustc_span::SessionGlobals>>::set::<rustc_span::create_session_if_not_set_then<core::result::Result<rustfmt_nightly::FormatReport, rustfmt_nightly::ErrorKind>, <rustfmt_nightly::Session<std::io::stdio::Stdout>>::format_input_inner::{closure#0}>::{closure#0}, core::result::Result<rustfmt_nightly::FormatReport, rustfmt_nightly::ErrorKind>>
  22: <rustfmt_nightly::Session<std::io::stdio::Stdout>>::format
  23: rustfmt::format_and_emit_report::<std::io::stdio::Stdout>
  24: <rustfmt_nightly::Session<std::io::stdio::Stdout>>::override_config::<rustfmt::format::{closure#0}, ()>
  25: rustfmt::execute
  26: rustfmt::main
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.

[WaffleLapkin]

Seems to be introduced in nightly-2022-01-18:

; cargo +nightly-2022-01-17 fmt && echo "(just fine)"
(just fine)
; cargo +nightly-2022-01-18 fmt                      
thread 'main' panicked at 'no errors encountered even though `delay_span_bug` issued', compiler/rustc_errors/src/lib.rs:1188:13
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

[calebcartwright]

Thanks for the report! A compiler panic is probably best reported in rust-lang/rust, even though it's being surfaced by a dev tool. I'll check with infra to see if it makes sense to move the issue

We actually haven't made any changes to rustfmt in that window and the minimal changes that were applied by folks working on the compiler were highly minimal and certainly don't strike me as anything that would be the root cause.

As such I suspect this is due to changes made elsewhere within the compiler mods, and would at least like for those folks to be aware (even if we need to make changes on the rustfmt side to address/work around)

Would you be up for trying to boil it down to a more simple repro case?

[WaffleLapkin]

Sure! Here is a MRE:

macro_rules! a { () => { A<'a,> }; }

; bat ./tt.rs --style="plain"
macro_rules! a { () => { A<'a,> }; }
; rustfmt +nightly-2022-01-18 ./tt.rs
thread 'main' panicked at 'no errors encountered even though `delay_span_bug` issued', compiler/rustc_errors/src/lib.rs:1188:13
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

[ehuss]

I have bisected to #92876. cc @compiler-errors

@calebcartwright I noticed that rustfmt is exiting with 0 even though it panicked. Is that normal?

[compiler-errors]

I'll look into it today or tomorrow!

@rustbot claim

[calebcartwright]

@calebcartwright I noticed that rustfmt is exiting with 0 even though it panicked. Is that normal?

Outside a macro, no, (as was the case in the original full repro), but yes within the context of a macro as in the minimal rerpo. We have our calls to rustc_parse internals wrapped in catch_unwind so the panic isn't really surfacing here, but rustfmt is still displaying the underlying error text prior to emitting formatting.

This is done partly due to the fact that there are scenarios where those internals panic that we want to wrap to an error type, and partly due to there being some scenarios we want to ignore for the purposes of formatting due to how we attempt to deal with macros in the pre-expansion world where rustfmt lives. The latter being a result of limited macro support and leaving macros it can't handle alone, while carrying on formatting the rest of the file/crate/workspace

[Samzyre]

It looks like I'm still getting this error with
rustc 1.60.0-nightly (2681f253b 2022-02-01)
rustfmt 1.4.38-nightly (2681f253 2022-02-01)

(Coming from rust-lang/rustfmt#5208 that was closed due to this one)

Playground seems to produce it also:
https://play.rust-lang.org/?version=nightly&mode=debug&edition=2021&gist=28266fbea1b847f180f6e3adbc99e4f4