fix: correct negative axis handling in roll function

[f14XuanLv]

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.

Description

This PR fixes a bug in xt::roll(e, shift, axis) where negative axis indices (e.g., -1 for the last axis) were incorrectly rejected.

The Bug

include/xtensor/misc/xmanipulation.hpp

The original code converted axis to size_t before normalization:

std::size_t saxis = static_cast<std::size_t>(axis);  // -1 becomes SIZE_MAX
if (axis < 0)
{
    axis += std::ptrdiff_t(cpy. dimension());
}

if (saxis >= cpy.dimension() || axis < 0)  // SIZE_MAX >= dim is always true! 
{
    XTENSOR_THROW(... );
}

This caused valid negative indices like -1 to incorrectly trigger the bounds check exception.

The Fix

include/xtensor/misc/xmanipulation.hpp

     auto cpy = empty_like(e);
     const auto& shape = cpy. shape();
-    std::size_t saxis = static_cast<std::size_t>(axis);
-    if (axis < 0)
-    {
-        axis += std::ptrdiff_t(cpy. dimension());
-    }
+    const auto dim = cpy.dimension();

-    if (saxis >= cpy.dimension() || axis < 0)
+    if (axis < -static_cast<std::ptrdiff_t>(dim) || axis >= static_cast<std::ptrdiff_t>(dim))
     {
-        XTENSOR_THROW(std::runtime_error, "axis is no within shape dimension.");
+        XTENSOR_THROW(std::runtime_error, "axis is not within shape dimension.");
     }

+    std::size_t saxis = normalize_axis(dim, axis);
+
     const auto axis_dim = static_cast<std::ptrdiff_t>(shape[saxis]);

• Validate axis bounds before conversion
• Use normalize_axis() for consistency with other functions (swapaxes, moveaxis, etc.)
• Fix typo: "axis is no within" → "axis is not within"

Tests Added

test/test_xmanipulation.cpp

     xarray<double> expected8 = {{{3, 1, 2}}, {{6, 4, 5}}, {{9, 7, 8}}};
     ASSERT_EQ(expected8, xt::roll(e2, -2, /*axis*/ 2));

+    // Boundary error cases
+    EXPECT_THROW(xt::roll(e2, 1, /*axis*/ 3), std::runtime_error);
+    EXPECT_THROW(xt::roll(e2, 1, /*axis*/ -4), std::runtime_error);
+
+    // Negative axis indices
+    xarray<double> expected9 = {{{3, 1, 2}}, {{6, 4, 5}}, {{9, 7, 8}}};
+    ASSERT_EQ(expected9, xt::roll(e2, -2, /*axis*/ -1));
+
+    xarray<double> expected10 = {{{1, 2, 3}}, {{4, 5, 6}}, {{7, 8, 9}}};
+    ASSERT_EQ(expected10, xt::roll(e2, -2, /*axis*/ -2));
+
+    xarray<double> expected11 = {{{4, 5, 6}}, {{7, 8, 9}}, {{1, 2, 3}}};
+    ASSERT_EQ(expected11, xt::roll(e2, 2, /*axis*/ -3));
 }

Note: This bug has existed since #1823 (2019).

[f14XuanLv]

Hi @JohanMabille 👋

This library is fantastic! I've been using xtensor in my projects and it works really well for multi-dimensional array operations.

I noticed an issue with negative axis handling in the roll function and submitted this PR to fix it.

Would appreciate if you could review when you have time. I also have another related PR in #2877 that adds multi-axis support for xt::roll. It would be helpful to prioritize this PR (#2878) first, so I can resolve any conflicts and adjust #2877 accordingly.

Thanks!