Make the BUG_REPORT_URL configurable for tools

[jyn514]

Right now, install_ice_hook hard-codes the bug report url to "https://github.com/rust-lang/rust/issues/new?labels=C-bug%2C+I-ICE%2C+T-compiler&template=ice.md":

rust/compiler/rustc_driver_impl/src/lib.rs

Line 1192 in 6502613

report_ice(info, BUG_REPORT_URL);

For tools, this is not correct; rustdoc at least wants T-rustdoc instead of T-compiler, and other tools want different repositories altogether. I'm opening this bug here because it's relevant to all tools, even though the original context was for clippy.

Fixing this seems a little tricky since DEFAULT_HOOK can't take an argument and doesn't have access to TyCtxt, but maybe we can solve this with thread-locals or something similar?

cc @rust-lang/rustdoc @rust-lang/clippy @rust-lang/miri @rust-lang/rustfmt
_Originally posted by @ingomancer in rust-lang/rust-clippy#10529

[flip1995]

Clippy already sets its own URL, so that is already possible. However Clippy needs to do a lot to do that:

rust/src/tools/clippy/src/driver.rs

Lines 196 to 252 in 439292b

	const BUG_REPORT_URL: &str = "https://github.com/rust-lang/rust-clippy/issues/new";
	type PanicCallback = dyn Fn(&panic::PanicInfo<'_>) + Sync + Send + 'static;
	static ICE_HOOK: LazyLock<Box<PanicCallback>> = LazyLock::new(|| {
	let hook = panic::take_hook();
	panic::set_hook(Box::new(|info| report_clippy_ice(info, BUG_REPORT_URL)));
	hook
	});
	fn report_clippy_ice(info: &panic::PanicInfo<'_>, bug_report_url: &str) {
	// Invoke our ICE handler, which prints the actual panic message and optionally a backtrace
	(*ICE_HOOK)(info);
	// Separate the output with an empty line
	eprintln!();
	let fallback_bundle = rustc_errors::fallback_fluent_bundle(rustc_driver::DEFAULT_LOCALE_RESOURCES.to_vec(), false);
	let emitter = Box::new(rustc_errors::emitter::EmitterWriter::stderr(
	rustc_errors::ColorConfig::Auto,
	None,
	None,
	fallback_bundle,
	false,
	false,
	None,
	false,
	false,
	rustc_errors::TerminalUrl::No,
	));
	let handler = rustc_errors::Handler::with_emitter(true, None, emitter);
	// a .span_bug or .bug call has already printed what
	// it wants to print.
	if !info.payload().is::<rustc_errors::ExplicitBug>() {
	let mut d = rustc_errors::Diagnostic::new(rustc_errors::Level::Bug, "unexpected panic");
	handler.emit_diagnostic(&mut d);
	}
	let version_info = rustc_tools_util::get_version_info!();
	let xs: Vec<Cow<'static, str>> = vec![
	"the compiler unexpectedly panicked. this is a bug.".into(),
	format!("we would appreciate a bug report: {bug_report_url}").into(),
	format!("Clippy version: {version_info}").into(),
	];
	for note in &xs {
	handler.note_without_error(note.as_ref());
	}
	// If backtraces are enabled, also print the query stack
	let backtrace = env::var_os("RUST_BACKTRACE").map_or(false, |x| &x != "0");
	let num_frames = if backtrace { None } else { Some(2) };
	interface::try_print_query_stack(&handler, num_frames);
	}

Definitely worth improving the situation here.

[jyn514]

Clippy already sets its own URL, so that is already possible. However Clippy needs to do a lot to do that:

Ah hmm, looks like the original problem was a bug in rustfix:

after fixes were automatically applied the compiler reported errors within these files:

  * src/main.rs

This likely indicates a bug in either rustc or cargo itself,
and we would appreciate a bug report! You're likely to see
a number of compiler warnings after this message which cargo
attempted to fix but failed. If you could open an issue at
https://github.com/rust-lang/rust/issues
quoting the full output of this command we'd be very appreciative!

I'll open a separate issue there.

[jyn514]

@rust-lang/miri the way I'm implementing this requires a change to the miri subtree in rust-lang/rust. What URL would you prefer for the bug report, just https://github.com/rust-lang/miri/issues/new ?

[jyn514]

#110989 - looks like rustfmt isn't using install_ice_hook today? Happy to add it, but I'll make the PR to the rust-lang/rustfmt repo instead of doing it here.

[saethlin]

In my experience, ICEs in Miri are just as often rustc bugs. I don't care if users report them in either place.

[calebcartwright]

#110989 - looks like rustfmt isn't using install_ice_hook today? Happy to add it, but I'll make the PR to the rust-lang/rustfmt repo instead of doing it here.

Do you have a ballpark feel for the size of the change? FWIW I'd be okay with making the rustfmt update in tree if it's reasonably small and logically part of a batch of related in-tree changes

[jyn514]

about 3 lines of code :) most of which will be the URL itself

[calebcartwright]

about 3 lines of code :) most of which will be the URL itself

sgtm, feel free to go ahead 👍