[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

pad and tile index their argument by axis, but only checked its size with XTENSOR_ASSERT, which is off in release builds. The limits of each pad mode were not checked in release either:

xt::xtensor<int, 2> a = {{1, 2, 3}, {4, 5, 6}};
xt::pad(a, std::vector<std::vector<std::size_t>>{{1, 1}});  // heap-buffer-overflow
xt::pad(a, {{5, 5}, {1, 1}}, xt::pad_mode::wrap);           // uninitialized values
xt::xtensor<int, 1> b = {1, 2, 3};
xt::tile(b, {2u, 3u});                                      // stack-buffer-overflow
xt::xarray<int> s = 5;
xt::tile(s, 3u);                                            // null dereference

Too many pairs were silently ignored, reflect with a width of at least the axis size and edge on an empty axis returned uninitialized values, and negative widths or repetitions wrapped to huge sizes.

These cases now throw std::runtime_error before the result is allocated. The loops in tile use the dimension of the expression for every buffer they index, which also removes the GCC 13 -O3 -Wstringop-overflow warning on its slice vector.

xpad.invalid_pad_width and xpad.tile_invalid_reps fail on master under AddressSanitizer and pass with the fix. xpad.largest_pad_width checks that the largest valid width of each mode still works.

[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-pad-tile-validation (4dccb74) with master (57f3d07)}