What does compare-mode=nll do these days?

[guswynn]

See #76765 (comment)

what does compare-mode=nll actually represent these days? I thought nll was long-since stable

cc @tmandry

[guswynn]

@rustbot modify labels: +A-diagnostics +T-compiler

[bjorn3]

I could find two differences:

$ rg "BorrowckMode::Mir|\.migrate|suppress_errors" compiler
compiler/rustc_middle/src/ty/context.rs
1336:        self.borrowck_mode().migrate()
1365:            return BorrowckMode::Mir;

compiler/rustc_infer/src/infer/lexical_region_resolve/mod.rs
45:        RegionckMode::Erase { suppress_errors: false } => {
56:        RegionckMode::Erase { suppress_errors: true } => {

compiler/rustc_infer/src/infer/mod.rs
105:        suppress_errors: bool,
124:            BorrowckMode::Migrate => RegionckMode::Erase { suppress_errors: false },
127:            BorrowckMode::Mir => RegionckMode::Erase { suppress_errors: true },

compiler/rustc_mir/src/borrow_check/mod.rs
1118:                ) if { tcx.migrate_borrowck() && this.borrow_set.contains(&location) } => {

compiler/rustc_data_structures/src/sso/map.rs
528:        self.ssomap.migrate_if_full();

compiler/rustc_session/src/config.rs
483:            BorrowckMode::Mir => false,
1568:        "mir" => BorrowckMode::Mir,

First regionck being skipped instead of given the opportunity to report errors:

rust/compiler/rustc_infer/src/infer/lexical_region_resolve/mod.rs

Lines 45 to 59 in 85fbf49

	RegionckMode::Erase { suppress_errors: false } => {
	// Do real inference to get errors, then erase the results.
	let mut values = resolver.infer_variable_values(&mut errors);
	let re_erased = region_rels.tcx.lifetimes.re_erased;
	values.values.iter_mut().for_each(|v| match *v {
	VarValue::Value(ref mut r) => *r = re_erased,
	VarValue::ErrorValue => {}
	});
	(values, errors)
	}
	RegionckMode::Erase { suppress_errors: true } => {
	// Skip region inference entirely.
	(resolver.erased_data(region_rels.tcx), Vec::new())
	}

Second a fix for #56254 ([NLL] prohibit "two-phase borrows" with existing borrows?)

rust/compiler/rustc_mir/src/borrow_check/mod.rs

Lines 1118 to 1134 in c4fe25d

	) if { tcx.migrate_borrowck() && this.borrow_set.contains(&location) } => {
	let bi = this.borrow_set.get_index_of(&location).unwrap();
	debug!(
	"recording invalid reservation of place: {:?} with \
	borrow index {:?} as warning",
	place_span.0, bi,
	);
	// rust-lang/rust#56254 - This was previously permitted on
	// the 2018 edition so we emit it as a warning. We buffer
	// these sepately so that we only emit a warning if borrow
	// checking was otherwise successful.
	this.reservation_warnings
	.insert(bi, (place_span.0, place_span.1, location, bk, borrow.clone()));
	// Don't suppress actual errors.
	Control::Continue
	}

[guswynn]

@bjorn3 interesting, in #76765, src/test/ui/issues/issue-76547.nll.stderr is materially different (and worse!) output in 'mir' instead of 'migrate'. I am inclined to think that, because nll is quite old at this point, and migrate is the default, we should perhaps get rid of the 2 modes and just use whatever codepaths are triggered by Migrate?(well, compare-mode=polonious usually sets -Zpolonious AND -borrowck=mir so perhaps we can just trigger the non-migrate paths with -Zpolonius?)

[bjorn3]

Regionck is an artifact of the fact that the old ast borrowck used lexical lifetimes. With polonius it doesn't even make sense to talk about regions anymore.

[guswynn]

@bjorn3 should we get rid of the distinction between "migrate" and "nll", it may say ci time and also will eventually be entirely replaced by polonious, as far as i can tell?

[bjorn3]

Migrate is the current behavior and nll is what it should become I think.

[jackh726]

Going to close this since NLL is tracked elsewhere.