--message-format short triggers internal cargo-miri panic

[kadiwa4]

Running cargo +nightly miri test --message-format short results in an internal cargo-miri panic.

Example lib.rs:

//! ```
//! assert_eq!(1 + 1, 2);
//! ```

Output

Preparing a sysroot for Miri (target: x86_64-apple-darwin)... done
    Finished test [unoptimized + debuginfo] target(s) in 0.00s
     Running unittests src/lib.rs (target/miri/x86_64-apple-darwin/debug/deps/tmp-320a0fd09b613afb)

running 0 tests

test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out

   Doc-tests tmp

running 1 test
test src/lib.rs - (line 1) ... FAILED

failures:

---- src/lib.rs - (line 1) stdout ----
Test executable failed (exit status: 101).

stderr:
thread 'main' panicked at src/tools/miri/cargo-miri/src/phases.rs:532:13:
assertion failed: suffix.starts_with(\'=\')
stack backtrace:
   0:        0x101a4c9ab - <std::sys_common::backtrace::_print::DisplayBacktrace as core::fmt::Display>::fmt::hc6e24405ab49cf6a
   1:        0x101a6e81b - core::fmt::write::hba1074a1739ff437
   2:        0x101a492ce - std::io::Write::write_fmt::h23c1134022d95821
   3:        0x101a4c789 - std::sys_common::backtrace::print::h10e3e1f9930ad6b0
   4:        0x101a4e4a8 - std::panicking::panic_hook_with_disk_dump::{{closure}}::heba4da2e45908525
   5:        0x101a4e1b5 - std::panicking::panic_hook_with_disk_dump::h6ca090a7ebe70b0e
   6:        0x101a4eac6 - std::panicking::rust_panic_with_hook::heb109e5af3a57c46
   7:        0x101a4e997 - std::panicking::begin_panic_handler::{{closure}}::h406476c3b1a5ecca
   8:        0x101a4ce89 - std::sys_common::backtrace::__rust_end_short_backtrace::h5966efc07d05eecf
   9:        0x101a4e6fd - _rust_begin_unwind
  10:        0x101a7c4a5 - core::panicking::panic_fmt::h4d90734c99616f06
  11:        0x101a7c537 - core::panicking::panic::h1950c2a470d70e81
  12:        0x1019cabc0 - cargo_miri[37c3d5d5e0cd7b06]::phases::phase_runner::<std[104a5f197c5359f5]::env::Args>
  13:        0x1019e0941 - cargo_miri[37c3d5d5e0cd7b06]::main
  14:        0x1019df7b6 - std[104a5f197c5359f5]::sys_common::backtrace::__rust_begin_short_backtrace::<fn(), ()>
  15:        0x1019e07cc - std[104a5f197c5359f5]::rt::lang_start::<()>::{closure#0}
  16:        0x101a4398a - std::rt::lang_start_internal::h476b5e1d4a1f36f0
  17:        0x1019e0c9c - _main



failures:
    src/lib.rs - (line 1)

test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.05s

error: doctest failed, to rerun pass `--doc`

I don't know if --message-format is supposed to be supported at all or not, but the message is not very nice. Feel free to close this if it doesn't feel important enough.

[RalfJung]

Ugh yeah that is ugly.

Out of curiosity, does cargo +nightly miri test --message-format=short work?

[kadiwa4]

No, same result.

[RalfJung]

That is strange, but maybe cargo parses and re-serializes the argument.

The issue is that here we only support arguments with the value passed via =, not via separate arguments for flag and value. This should be fairly easy to fix: if the suffix is empty, also drop the next argument.

[RalfJung]

Though thinking about it, that might just show up as the next bug where we don't respect --mesage-format for doctests. Looks like we should maybe just not mess with --message-format and --json when phase is Rustdoc? Not sure if the --extern patching is still needed, but it probably is.