Hello, and thank you for maintaining this crate.
I found a sanitizer/Miri failure reachable through public crate APIs using safe Rust code. I may be missing crate-specific preconditions, but the behavior looks worth checking because safe callers should not be able to trigger undefined behavior.
Summary
- Crate:
peroxide
- Version tested:
0.41.0
- API paths with similar failures observed:
MatrixTrait::subs_col
MatrixTrait::subs_row
- Matrix multiplication:
Matrix * Matrix, &Matrix * &Matrix
Matrix indexing/mutation paths
- Linear algebra operations over public
Matrix state
Observed diagnostics
Miri undefined behavior: error: Undefined Behavior: overflowing pointer arithmetic: the total offset in bytes does not fit in an `isize`
Unsafe precondition violation: unsafe precondition(s) violated: ptr::add requires that the address calculation does not overflow
Reproduction
The snippets below are minimal readable reproducers. Each PoC also includes the source location most relevant to the reported failure.
PoC 1: test_peroxide11::generated_test_11
Relevant source location:
src/structure/matrix.rs:701-705: Matrix exposes data, row, col, and shape as public fields.
src/structure/matrix.rs:1148: subs_col operates on those unchecked shape/storage invariants.
Readable equivalent PoC:
use peroxide::fuga::{Matrix, MatrixTrait, Shape};
#[test]
fn poc() {
let mut rhs = Matrix {
data: Vec::new(),
row: usize::MAX,
col: usize::MAX,
shape: Shape::Row,
};
rhs.subs_col(usize::MAX, &[]);
let lhs = Matrix {
data: Vec::new(),
row: usize::MAX,
col: 0,
shape: Shape::Col,
};
let _ = lhs * rhs;
}
Observed diagnostic:
error: Undefined Behavior: overflowing pointer arithmetic: the total offset in bytes does not fit in an `isize`
PoC 2: test_peroxide12::generated_test_12
Relevant source location:
src/structure/matrix.rs:701-705: Matrix exposes data, row, col, and shape as public fields.
src/structure/matrix.rs:1156: subs_row operates on those unchecked shape/storage invariants.
Readable equivalent PoC:
use peroxide::fuga::{Matrix, MatrixTrait, Shape};
#[test]
fn poc() {
let mut rhs = Matrix {
data: Vec::new(),
row: 0,
col: usize::MAX,
shape: Shape::Col,
};
rhs.subs_row(usize::MAX, &[]);
let lhs = Matrix {
data: Vec::new(),
row: usize::MAX,
col: usize::MAX,
shape: Shape::Row,
};
let _ = &lhs * &rhs;
}
Observed diagnostic:
unsafe precondition(s) violated: ptr::add requires that the address calculation does not overflow
Source review and suggested fix
Brief reasoning:
Matrix exposes public fields, so safe code can construct states that do not satisfy the library invariants.
- The later indexing and arithmetic code assumes those invariants without rechecking them.
Suggested fix:
- Hide the raw fields or validate shape and storage invariants before any unsafe indexing or pointer arithmetic.
Thanks again for taking a look.
Hello, and thank you for maintaining this crate.
I found a sanitizer/Miri failure reachable through public crate APIs using safe Rust code. I may be missing crate-specific preconditions, but the behavior looks worth checking because safe callers should not be able to trigger undefined behavior.
Summary
peroxide0.41.0MatrixTrait::subs_colMatrixTrait::subs_rowMatrix * Matrix,&Matrix * &MatrixMatrix indexing/mutation pathsMatrixstateObserved diagnostics
Reproduction
The snippets below are minimal readable reproducers. Each PoC also includes the source location most relevant to the reported failure.
PoC 1:
test_peroxide11::generated_test_11Relevant source location:
src/structure/matrix.rs:701-705:Matrixexposesdata,row,col, andshapeas public fields.src/structure/matrix.rs:1148:subs_coloperates on those unchecked shape/storage invariants.Readable equivalent PoC:
Observed diagnostic:
PoC 2:
test_peroxide12::generated_test_12Relevant source location:
src/structure/matrix.rs:701-705:Matrixexposesdata,row,col, andshapeas public fields.src/structure/matrix.rs:1156:subs_rowoperates on those unchecked shape/storage invariants.Readable equivalent PoC:
Observed diagnostic:
Source review and suggested fix
Brief reasoning:
Matrixexposes public fields, so safe code can construct states that do not satisfy the library invariants.Suggested fix:
Thanks again for taking a look.