CLI errors show escape codes

[epage]

Verification

• I searched for recent similar issues at https://github.com/rust-lang/rustup/issues?q=is%3Aissue and found no duplicates.
• I am on the latest version of Rustup according to https://github.com/rust-lang/rustup/tags and am still able to reproduce my issue.

Problem

For some CLI errors, rustup shows escape codes while others show colored output.

Steps

$ cargo run -- show --jkhjkh
warning: ignoring 'build.analysis' config, pass `-Zbuild-analysis` to enable it
    Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.09s
     Running `target/debug/rustup-init show --jkhjkh`
error: \x1b[1m\x1b[91merror:\x1b[0m unexpected argument '\x1b[1m\x1b[33mshow\x1b[0m' found

\x1b[1m\x1b[92mUsage:\x1b[0m \x1b[1m\x1b[96mrustup-init[EXE]\x1b[0m \x1b[36m[OPTIONS]\x1b[0m

For more information, try '\x1b[1m\x1b[96m--help\x1b[0m'.
$ cargo run --config env.RUSTUP_FORCE_ARG0=\'rustup\' -- check --no-self-update=foo
warning: ignoring 'build.analysis' config, pass `-Zbuild-analysis` to enable it
    Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.08s
     Running `target/debug/rustup-init check --no-self-update=foo`
error: unexpected value 'foo' for '--no-self-update' found; no more were expected

Usage: rustup[EXE] check --no-self-update

For more information, try '--help'.

Possible Solution(s)

No response

Notes

I'm assuming something is weird about the two streams being selected between at

rustup/src/cli/rustup_mode.rs

Lines 613 to 624 in a373a2c

	if [
	InvalidSubcommand,
	UnknownArgument,
	DisplayHelpOnMissingArgumentOrSubcommand,
	]
	.contains(&err.kind())
	{
	write!(process.stdout().lock(), "{}", err.render().ansi())?;
	} else {
	write!(process.stderr().lock(), "{}", err.render().ansi())?;
	}
	return Ok(ExitCode::FAILURE);

Rustup version

$ cargo run -- --version
warning: ignoring 'build.analysis' config, pass `-Zbuild-analysis` to enable it
    Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.10s
     Running `target/debug/rustup-init --version`
rustup-init 1.29.0 :: 1.28.2+518 (a373a2c8c 2026-02-24)

Installed toolchains

N/A

OS version

N/A

[epage]

I was wrong, I think the issue is related to

rustup/src/cli/setup_mode.rs

Lines 101 to 106 in a373a2c

	Err(e) if [ErrorKind::DisplayHelp, ErrorKind::DisplayVersion].contains(&e.kind()) => {
	use std::io::Write as _;
	write!(process.stdout().lock(), "{}", e.render().ansi())?;
	return Ok(utils::ExitCode::SUCCESS);
	}
	Err(e) => return Err(format_err!("{}", e.render().ansi())),

I had thought I was able to reproduce this in rustup mode but I'm only seeing it in setup mode.

[epage]

rustup/src/cli/setup_mode.rs

Lines 101 to 106 in a373a2c

	Err(e) if [ErrorKind::DisplayHelp, ErrorKind::DisplayVersion].contains(&e.kind()) => {
	use std::io::Write as _;
	write!(process.stdout().lock(), "{}", e.render().ansi())?;
	return Ok(utils::ExitCode::SUCCESS);
	}
	Err(e) => return Err(format_err!("{}", e.render().ansi())),

rustup/src/cli/common.rs

Lines 461 to 472 in a373a2c

	pub fn report_error(e: &anyhow::Error, process: &Process) {
	// NB: This shows one error: even for multiple causes and backtraces etc,
	// rather than one per cause, and one for the backtrace. This seems like a
	// reasonable tradeoff, but if we want to do differently, this is the code
	// hunk to revisit, that and a similar build.rs auto-detect glue as anyhow
	// has to detect when backtrace is available.
	if show_backtrace(process) {
	error!("{:?}", e);
	} else {
	error!("{:#}", e);
	}
	}

tracing-subscriber used to pass along ANSI escape codes but just in case someone was logging untrusted data, they started escaping them (not even stripping them) to avoid people changing the title of your terminal?

See tokio-rs/tracing#3368

[epage]

In that PR

The fix specifically targets untrusted user input while preserving the ability for applications to deliberately include formatting in trusted contexts like thread names.

but no answer was given when asked

tokio-rs/tracing#3369 was later created with no update from the maintainers.

Options:

• Pin to an older tracing-subscriber (0.3.19)
• Not log this message but directly print it using anstream like the other error messages in rustup mode
• Find a way to swap out tracing-subscriber or at least the part that is doing the escaping

Opinions?

[rami3l]

@epage Sorry for the late reply, and thanks again for your thorough investigation!

I guess for now we can pin trancing-subscriber to begin with...

But in the end, if the impact of the said changes is limited, maybe we can go with your second option and use an alternative approach in sections just like report_error()?

I do hope they can add an unsafe option to at least let us go with certain escape codes 🤔