Start emitting overflow lints for const promoted SHL & SHRs #119339
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This fixes the issue where no lints were firing for overflowing SHL and SHR ops. That happened because `ConstPropagator::check_binary_op()`, which is responsible for firing these lints, was skipping them because it couldn't find any `Shl()` or `Shr()` ops in the main MIR. That in turn occurred because our const promotion pass `PromoteTemps` removes these ops from the MIR when it runs. With this change, the firing of these particular lints is predicated on the presence of an assert terminator instead, which fortunately is not removed by `PromoteTemps`
|
r? @b-naber (rustbot has picked a reviewer for you, use r? to override) |
rustbot
added
S-waiting-on-review
|
@rustbot author |
rustbot
added
S-waiting-on-author
This comment has been minimized.
This comment has been minimized.
RalfJung
reviewed
Dec 28, 2023
| // but it can't because they are getting removed from the MIR | ||
| // during const promotion. So we check the associated asserts | ||
| // here instead as they are not removed by promotion. | ||
| (AssertKind::Overflow(*bin_op, eval_to_int(op1), eval_to_int(op2)), true) |
There was a problem hiding this comment.
If you emit the lint here then you should be able to remove the shift handling from check_binary_op. That should also fix the duplicate lints.
There was a problem hiding this comment.
I tried that, but it eliminates only 6 of the 59 new duplicates added by this change.
There was a problem hiding this comment.
What exactly did you try? I should look something like this:
diff --git a/compiler/rustc_mir_transform/src/const_prop_lint.rs b/compiler/rustc_mir_transform/src/const_prop_lint.rs
index 99eecb567f2..d36798518e2 100644
--- a/compiler/rustc_mir_transform/src/const_prop_lint.rs
+++ b/compiler/rustc_mir_transform/src/const_prop_lint.rs
@@ -317,37 +317,9 @@ fn check_binary_op(
});
let l = self
.use_ecx(location, |this| this.ecx.read_immediate(&this.ecx.eval_operand(left, None)?));
- // Check for exceeding shifts *even if* we cannot evaluate the LHS.
+ // Shifts are linted via the `Assert` terminator, not this function.
if matches!(op, BinOp::Shr | BinOp::Shl) {
- let r = r.clone()?;
- // We need the type of the LHS. We cannot use `place_layout` as that is the type
- // of the result, which for checked binops is not the same!
- let left_ty = left.ty(self.local_decls(), self.tcx);
- let left_size = self.ecx.layout_of(left_ty).ok()?.size;
- let right_size = r.layout.size;
- let r_bits = r.to_scalar().to_bits(right_size).ok();
- if r_bits.is_some_and(|b| b >= left_size.bits() as u128) {
- debug!("check_binary_op: reporting assert for {:?}", location);
- let source_info = self.body().source_info(location);
- let panic = AssertKind::Overflow(
- op,
- match l {
- Some(l) => l.to_const_int(),
- // Invent a dummy value, the diagnostic ignores it anyway
- None => ConstInt::new(
- ScalarInt::try_from_uint(1_u8, left_size).unwrap(),
- left_ty.is_signed(),
- left_ty.is_ptr_sized_integral(),
- ),
- },
- r.to_const_int(),
- );
- self.report_assert_as_lint(
- source_info,
- AssertLint::ArithmeticOverflow(source_info.span, panic),
- );
- return None;
- }
+ return None;
}
if let (Some(l), Some(r)) = (l, r) {There was a problem hiding this comment.
Oh. I just got rid of the entire if block. Didn't realize I must return None to disable shift handling.
I have done that now in the latest commit I pushed. It has eliminated all duplicates. However, we have the opposite problem now. Many of the shift lints in release builds have been silenced possibly because we don't emit asserts in those builds.
There was a problem hiding this comment.
Many of the shift lints in release builds have been silenced possibly because we don't emit asserts in those builds.
Ah... right. Yeah for debug/release consistency we really want to lint on the operation (shl/shr/add/...), not the assert. But then we need to lint in promoteds as well.
|
The job Click to see the possible cause of the failure (guessed by this bot) |
|
☔ The latest upstream changes (presumably #119627) made this pull request unmergeable. Please resolve the merge conflicts. |
|
I think this got superseded by #119432? |
Yes. I'll abandon it. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fixes #117949
The overflow lints for promoted shift ops had been accidently disabled by #108282. That is because #108282 caused
Shl&Shrops to be removed from the MIR which silenced the lints.In this PR we start firing them again by triggering them based on the assert terminators instead which luckily are not removed from the MIR.