[gouarin]

Checklist

• The title and commit message(s) are descriptive.
• Small commits made to fix your PR have been squashed to avoid history pollution.
• Tests have been added for new features or bug fixes.
• API of new functions and classes are documented (no public API changes).

Description

The shape of an xtensor<T, N> is a std::array<std::size_t, N>. resize checks that a runtime shape has N values, but from_shape only had an XTENSOR_ASSERT and reshape had no check, so a shape of the wrong length was copied past the array:

xt::xtensor<double, 1>::from_shape(std::vector<std::size_t>{2, 3});  // stack-buffer-overflow
xt::xtensor<double, 1> t = {1., 2., 3., 4., 5., 6.};
t.reshape(std::vector<std::size_t>{2, 3});                           // same
xt::xarray<double> x = {{0.5, 1.5}, {2.5, 3.5}};
xt::interp(x, xp, fp);  // builds its xtensor<T, 1> result with from_shape(x.shape())

from_shape and both reshape overloads now use the same XTENSOR_PRECONDITION as resize. interp also checks its arguments up front: x, xp and fp must be 1-D, xp must not be empty, and fp must have the length of xp. Before, a 2-D xp or a short fp was read out of bounds, and a 0-D x gave a result with an uninitialized shape.

This removes the GCC 13/15 -Warray-bounds and -Wmaybe-uninitialized warnings on ret in xtensor.hpp (from_shape, reached from interp in test_extended_xmath_interp).

xtensor.change_dimension and xmath.interp_invalid_arguments fail on master and pass with the fix, also under AddressSanitizer and UBSan.

[codspeed]

Merging this PR will not alter performance

1 warning

⚠️ Unable to measure the execution of 1 benchmark

This benchmark measured no execution time, so its result is not comparable and counts as unchanged. This usually happens when the compiler removes the code under test.

Prevent compiler optimizations

See affected benchmarks

✅ 255 untouched benchmarks

---

_{Comparing gouarin:fix-xtensor-rank-checks (4836e24) with master (57f3d07)}