Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[InstCombine] Crash on valid input when narrowing type but not updating range attribute #112633

Closed
goldsteinn opened this issue Oct 16, 2024 · 1 comment · Fixed by #112649
Closed

[InstCombine] Crash on valid input when narrowing type but not updating range attribute #112633

goldsteinn opened this issue Oct 16, 2024 · 1 comment · Fixed by #112649
Assignees
@goldsteinn
Labels
crash-on-valid llvm:instcombine llvm:ir llvm:transforms

Comments

@goldsteinn
Copy link
Contributor

https://godbolt.org/z/b3M1dj6aj

define void @SetRow_C(ptr %p, i32 %conv) {
entry:
  %call.i = tail call ptr @__memset_chk(ptr %p, i32 range(i32 0, 123) %conv, i64 1, i64 1)
  ret void
}

declare ptr @__memset_chk(ptr, i32, i64, i64)

We keep the range(i32 ...) when we replace the _chk with memset.inline.p0.i64 which leads to a crash:

Range bit width must match type bit width!
  tail call void @llvm.memset.p0.i64(ptr noundef nonnull align 1 dereferenceable(1) %p, i8 range(i32 0, 250) %0, i64 1, i1 false)
@goldsteinn goldsteinn self-assigned this Oct 16, 2024
@goldsteinn
Copy link
Contributor Author

Ill have fix up in a bit. Going to basically add AttributeSet to typeIncompatible and drop range if it has become invalid.

@goldsteinn goldsteinn mentioned this issue Oct 17, 2024
Merged
goldsteinn added a commit to goldsteinn/llvm-project that referenced this issue Oct 17, 2024
@goldsteinn
[SimplifyLibCall][Attribute] Fix bug where we may keep range attr w…
d105ec7 
…ith incompatible type

In a variety of places we change the bitwidth of a parameter but don't
update the attributes.

The issue in this case is from the `range` attribute when inlining
`__memset_chk`. `optimizeMemSetChk` will replace an `i32` with an
`i8`, and if the `i32` had a `range` attr assosiated it will cause an
error.

Fixes llvm#112633
goldsteinn added a commit that referenced this issue Oct 17, 2024
@goldsteinn
[SimplifyLibCall][Attribute] Fix bug where we may keep range attr w…
c85611e 
…ith incompatible type (#112649)

In a variety of places we change the bitwidth of a parameter but don't
update the attributes.

The issue in this case is from the `range` attribute when inlining
`__memset_chk`. `optimizeMemSetChk` will replace an `i32` with an
`i8`, and if the `i32` had a `range` attr assosiated it will cause an
error.

Fixes #112633
@goldsteinn goldsteinn mentioned this issue Oct 17, 2024
Merged
goldsteinn added a commit to goldsteinn/llvm-project that referenced this issue Oct 17, 2024
@goldsteinn
Reapply "[Inliner] Propagate more attributes to params when inlining (l…
e64962e 
…lvm#91101)" (2nd Attempt)

Root cause of the bug was code hanging onto `range` attr after
changing BitWidth. This was fixed in PR llvm#112633.
goldsteinn added a commit that referenced this issue Oct 18, 2024
@goldsteinn
Reapply "[Inliner] Propagate more attributes to params when inlining (#…
69a798a 
…91101)" (2nd Attempt) (#112749)

Root cause of the bug was code hanging onto `range` attr after
changing BitWidth. This was fixed in PR #112633.
bricknerb pushed a commit to bricknerb/llvm-project that referenced this issue Oct 21, 2024
@goldsteinn @bricknerb
[SimplifyLibCall][Attribute] Fix bug where we may keep range attr w…
17c9801 
…ith incompatible type (llvm#112649)

In a variety of places we change the bitwidth of a parameter but don't
update the attributes.

The issue in this case is from the `range` attribute when inlining
`__memset_chk`. `optimizeMemSetChk` will replace an `i32` with an
`i8`, and if the `i32` had a `range` attr assosiated it will cause an
error.

Fixes llvm#112633
bricknerb pushed a commit to bricknerb/llvm-project that referenced this issue Oct 21, 2024
@goldsteinn @bricknerb
Reapply "[Inliner] Propagate more attributes to params when inlining (l…
34d473b 
…lvm#91101)" (2nd Attempt) (llvm#112749)

Root cause of the bug was code hanging onto `range` attr after
changing BitWidth. This was fixed in PR llvm#112633.
EricWF pushed a commit to efcs/llvm-project that referenced this issue Oct 22, 2024
@goldsteinn @EricWF
[SimplifyLibCall][Attribute] Fix bug where we may keep range attr w…
6cb2ff3 
…ith incompatible type (llvm#112649)

In a variety of places we change the bitwidth of a parameter but don't
update the attributes.

The issue in this case is from the `range` attribute when inlining
`__memset_chk`. `optimizeMemSetChk` will replace an `i32` with an
`i8`, and if the `i32` had a `range` attr assosiated it will cause an
error.

Fixes llvm#112633
EricWF pushed a commit to efcs/llvm-project that referenced this issue Oct 22, 2024
@goldsteinn @EricWF
Reapply "[Inliner] Propagate more attributes to params when inlining (l…
f2a2c86 
…lvm#91101)" (2nd Attempt) (llvm#112749)

Root cause of the bug was code hanging onto `range` attr after
changing BitWidth. This was fixed in PR llvm#112633.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@goldsteinn goldsteinn

Labels
crash-on-valid llvm:instcombine llvm:ir llvm:transforms
Projects
None yet
Milestone
No milestone
2 participants
@EugeneZelenko @goldsteinn