Fix ICE on invalid const param types

[gurry]

Fixes ICE #123863 which occurs because the const param has a type which is not a bool, char or an integral type.

The ICEing code path begins here in typeck_with_fallback:

rust/compiler/rustc_hir_typeck/src/lib.rs

Line 167 in cb3752d

let expected_type = expected_type.unwrap_or_else(fallback);

The fallback invokes the type_of query and that eventually ends up calling ct_infer from the lowering code over here:

rust/compiler/rustc_hir_analysis/src/hir_ty_lowering/mod.rs

Line 561 in cb3752d

self.lowerer.ct_infer(ty, Some(param), self.span).into()

and ct_infer ICEs at this location:

rust/compiler/rustc_hir_analysis/src/collect.rs

Line 392 in cb3752d

r => bug!("unexpected region: {r:?}"),

To fix the ICE it I'm triggering a span_delayed_bug before we hit ct_infer if the type of the const param is not one of the supported types

Edit

On @lcnr's suggestion I've changed the approach to not let ReStatic region hit the bug! in ct_infer instead of triggering a span_delayed_bug.

[rustbot]

r? @Nadrieril

rustbot has assigned @Nadrieril.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[gurry]

I need help with this fix.

It currently does not cater to the case when the feature adt_const_params is enabled because I am not sure how I should handle that. When we do a similar const param type check during WF we handle adt_const_param by running a proper WF check logic over here:

rust/compiler/rustc_hir_analysis/src/check/wfcheck.rs

Lines 938 to 954 in cb3752d

	if tcx.features().adt_const_params {
	enter_wf_checking_ctxt(tcx, hir_ty.span, param.def_id, |wfcx| {
	let trait_def_id =
	tcx.require_lang_item(LangItem::ConstParamTy, Some(hir_ty.span));
	wfcx.register_bound(
	ObligationCause::new(
	hir_ty.span,
	param.def_id,
	ObligationCauseCode::ConstParam(ty),
	),
	wfcx.param_env,
	ty,
	trait_def_id,
	);
	Ok(())
	})
	} else {

But we can't call WF from the location in this fix because that will lead to cycles (typeck -> type_of -> fix location -> wf -> typeck).

@Nadrieril Can you please suggest a solution to this conundrum?

[Nadrieril]

I'm afraid I know very little about typeck. Let's see if someone else can help

r? compiler

[lcnr]

please add the affected test to you PR

[lcnr]

i think the correct fix is to change

rust/compiler/rustc_hir_analysis/src/collect.rs

Lines 389 to 392 in 9adafa7

	// This is never reached in practice. If it ever is reached,
	// `ReErased` should be changed to `ReStatic`, and any other region
	// left alone.
	r => bug!("unexpected region: {r:?}"),

to keep 'static as is and not bug! there. Please try thatr

[gurry]

i think the correct fix is to change

rust/compiler/rustc_hir_analysis/src/collect.rs

Lines 389 to 392 in 9adafa7

	// This is never reached in practice. If it ever is reached,
	// `ReErased` should be changed to `ReStatic`, and any other region
	// left alone.
	r => bug!("unexpected region: {r:?}"),

to keep 'static as is and not bug! there. Please try thatr

Thanks @lcnr. I've implemented your suggestion and it has fixed the ICE.

[lcnr]

@bors r+ rollup

[bors]

📌 Commit c62bc31 has been approved by lcnr

It is now in the queue for this repository.