Fix: initialize facet to -1 in line_positive_intersect to prevent undefined behavior

[Dev-sh21]

hi @vissarion sir

In HPolytope::line_positive_intersect (the update_parameters overload, ~line 520 of hpolytope.h), the variable facet is declared but never initialized:

int m = num_of_hyperplanes(), facet;  // uninitialized

this by default takes garabge value, so it leads to error.
i have set this value to -1.

i have also attached all the proofs below.

[Mohit-Lakra]

LGTM

[Dev-sh21]

hi @vissarion sir, is there any change required?

[vissarion]

Thanks for this PR. Please provide the code triggered the bug. The screenshots are not that useful.

[Dev-sh21]

hi @vissarion sir,
apologies for the confusion

here is the minimal reproducible c++ snippet that triggers the bug, if a ray originates at the boundary and points directly outwards, it never finds a positive intersection (lamda > 0 condition fails).

// Create a simple 1D polytope: x <= 1
Eigen::MatrixXd A(1, 1); A << 1.0;
Eigen::VectorXd b(1); b << 1.0;
HPolytope<Cartesian<double>> P(1, A, b);

// Origin point on the boundary (x = 1), pointing outwards (+1 direction)
Point r(1); r.set_coord(0, 1.0);
Point v(1); v.set_coord(0, 1.0);

Eigen::VectorXd Ar, Av;
// This will return an uninitialized (garbage) integer for facet
// because no lambda > 0 is found.
std::pair<double, int> intersection = P.line_positive_intersect(r, v, Ar, Av);

[Dev-sh21]

in the [line_positive_intersect] function, the facet variable is declared without initialization
When a ray does not positively intersect the polytope (meaning lamda is never > 0), the (if (lamda < min_plus && lamda > 0)) block is never executed, this causes the function to return a garbage value

by explicitly initializing facet = -1 at declaration, we ensure that if no positive intersection is found, the function returns a deterministic invalid state -1 rather than triggering undefined behavior with a garbage value.

[vissarion]

in the [line_positive_intersect] function, the facet variable is declared without initialization When a ray does not positively intersect the polytope (meaning lamda is never > 0), the (if (lamda < min_plus && lamda > 0)) block is never executed, this causes the function to return a garbage value

by explicitly initializing facet = -1 at declaration, we ensure that if no positive intersection is found, the function returns a deterministic invalid state -1 rather than triggering undefined behavior with a garbage value.

Thank you for the example. Out of curiosity how you discovered that issue? Do you have a sampling or volume example that triggers this behavior?

[Dev-sh21]

in the [line_positive_intersect] function, the facet variable is declared without initialization When a ray does not positively intersect the polytope (meaning lamda is never > 0), the (if (lamda < min_plus && lamda > 0)) block is never executed, this causes the function to return a garbage value
by explicitly initializing facet = -1 at declaration, we ensure that if no positive intersection is found, the function returns a deterministic invalid state -1 rather than triggering undefined behavior with a garbage value.

Thank you for the example. Out of curiosity how you discovered that issue? Do you have a sampling or volume example that triggers this behavior?

hi @vissarion sir,
i found the uninitialized facet while running ghmc sampling on a 10d h-cube, where it caused occasional crashes during reflections. i've updated the pr to initialize facet to -1 and resolved gaussian walk signature mismatches to ensure all sampling tests pass locally