Also generate StorageDead in constants

[oli-obk]

r? @eddyb

None of this special casing is actually necessary since we started promoting within constants and statics.

We may want to keep some of it around out of perf reasons, but it's not required for user visible behaviour

somewhat related: #68622

[oli-obk]

cc @rust-lang/wg-const-eval

[oli-obk]

@bors try @rust-timer queue

[rust-timer]

Awaiting bors try build completion

[bors]

⌛ Trying commit 3f2f0dc383a6facaa1d64a21a8923d39943cc2ef with merge 0ccb4b9b8cc4437f530995e70c33db2fa684fc81...

[bors]

☀️ Try build successful - checks-actions
Build commit: 0ccb4b9b8cc4437f530995e70c33db2fa684fc81 (0ccb4b9b8cc4437f530995e70c33db2fa684fc81)

[rust-timer]

Queued 0ccb4b9b8cc4437f530995e70c33db2fa684fc81 with parent 338f939, future comparison URL.

[rust-timer]

Finished benchmarking try commit (0ccb4b9b8cc4437f530995e70c33db2fa684fc81): comparison url.

Benchmarking this pull request likely means that it is perf-sensitive, so we're automatically marking it as not fit for rolling up. Please note that if the perf results are neutral, you should likely undo the rollup=never given below by specifying rollup- to bors.

Importantly, though, if the results of this run are non-neutral do not roll this PR up -- it will mask other regressions or improvements in the roll up.

@bors rollup=never
@rustbot modify labels: +S-waiting-on-review -S-waiting-on-perf

[RalfJung]

Looks like we have a 5% regression in a stress test and a 1.5% regression on "ucd".

[eddyb]

This comment was copied from here (as part of #52405's c3dbdfe):

rust/src/librustc_mir/build/expr/as_temp.rs

Lines 61 to 68 in c3dbdfe

	// In constants, temp_lifetime is None. We should not need to drop
	// anything because no values with a destructor can be created in
	// a constant at this time, even if the type may need dropping.
	ifletSome(temp_lifetime) = temp_lifetime {
	this.schedule_drop_storage_and_value(
	expr_span, temp_lifetime,&Place::Local(temp), expr_ty,
	);
	}

This is the current version, updated in (your) #56127:

rust/compiler/rustc_mir_build/src/build/expr/as_temp.rs

Lines 98 to 113 in e0ef0fc

	// In constants, `temp_lifetime` is `None` for temporaries that
	// live for the `'static` lifetime. Thus we do not drop these
	// temporaries and simply leak them.
	// This is equivalent to what `let x = &foo();` does in
	// functions. The temporary is lifted to their surrounding
	// scope. In a function that means the temporary lives until
	// just before the function returns. In constants that means it
	// outlives the constant's initialization value computation.
	// Anything outliving a constant must have the `'static`
	// lifetime and live forever.
	// Anything with a shorter lifetime (e.g the `&foo()` in
	// `bar(&foo())` or anything within a block will keep the
	// regular drops just like runtime code.
	ifletSome(temp_lifetime) = temp_lifetime {
	this.schedule_drop(expr_span, temp_lifetime, temp,DropKind::Storage);
	}

If we don't want to copy the comment (which is pretty long), we could at least point to the other one?
Or if that's not applicable, it could still mention as_temp but say that the same situation does not arise for what's handled here.

[eddyb]

I think this is strictly "more correct", even if less efficient. r=me if you want to land it as-is (modulo that one comment) but it might be prudent to wait for more opinions from @rust-lang/compiler.

[nikomatsakis]

I'm in favor of removing the special casing

[oli-obk]

@bors r=eddyb

[bors]

📌 Commit 84fe7cf has been approved by eddyb

[bors]

⌛ Testing commit 84fe7cf with merge cc03ee6...

[bors]

☀️ Test successful - checks-actions
Approved by: eddyb
Pushing cc03ee6 to master...