BUG: Fix zero-padded copy when truncating (ticket #1756)

Chuck's fix was correct, but there was another bug at play here.

I've also removed the _strided_to_contig and _contig_to_strided
specializations as they likely don't provide a significant advantage
over _strided_to_strided here.

2 files changed, 37 insertions, 42 deletions

diff --git a/numpy/core/src/multiarray/dtype_transfer.c b/numpy/core/src/multiarray/dtype_transfer.c
index ee0ec720d..b64e1684a 100644
--- a/numpy/core/src/multiarray/dtype_transfer.c
+++ b/numpy/core/src/multiarray/dtype_transfer.c
@@ -185,6 +185,27 @@ _strided_to_strided_zero_pad_copy(char *dst, npy_intp dst_stride,
     }
 }
 
+/*
+ * Does a strided to strided zero-padded copy for the case where
+ * dst_itemsize < src_itemsize
+ */
+static void
+_strided_to_strided_truncate_copy(char *dst, npy_intp dst_stride,
+                        char *src, npy_intp src_stride,
+                        npy_intp N, npy_intp src_itemsize,
+                        void *data)
+{
+    _strided_zero_pad_data *d = (_strided_zero_pad_data *)data;
+    npy_intp dst_itemsize = d->dst_itemsize;
+
+    while (N > 0) {
+        memcpy(dst, src, dst_itemsize);
+        src += src_stride;
+        dst += dst_stride;
+        --N;
+    }
+}
+
 NPY_NO_EXPORT int
 PyArray_GetStridedZeroPadCopyFn(int aligned,
                             npy_intp src_stride, npy_intp dst_stride,
@@ -192,13 +213,9 @@ PyArray_GetStridedZeroPadCopyFn(int aligned,
                             PyArray_StridedTransferFn **out_stransfer,
                             void **out_transferdata)
 {
-    if (src_itemsize >= dst_itemsize) {
-        /* If the sizes are different, the alignment flag isn't trustworthy */
-        if (src_itemsize != dst_itemsize) {
-            aligned = 0;
-        }
+    if (src_itemsize == dst_itemsize) {
         *out_stransfer = PyArray_GetStridedCopyFn(aligned, src_stride,
-                                dst_stride, dst_itemsize);
+                                dst_stride, src_itemsize);
         *out_transferdata = NULL;
         return (*out_stransfer == NULL) ? NPY_FAIL : NPY_SUCCEED;
     }
@@ -213,7 +230,13 @@ PyArray_GetStridedZeroPadCopyFn(int aligned,
         d->freefunc = &PyArray_free;
         d->copyfunc = &_strided_zero_pad_data_copy;
 
-        *out_stransfer = &_strided_to_strided_zero_pad_copy;
+        if (src_itemsize < dst_itemsize) {
+            *out_stransfer = &_strided_to_strided_zero_pad_copy;
+        }
+        else {
+            *out_stransfer = &_strided_to_strided_truncate_copy;
+        }
+
         *out_transferdata = d;
         return NPY_SUCCEED;
     }
diff --git a/numpy/core/src/multiarray/lowlevel_strided_loops.c.src b/numpy/core/src/multiarray/lowlevel_strided_loops.c.src
index c5f408d81..fc8d71f46 100644
--- a/numpy/core/src/multiarray/lowlevel_strided_loops.c.src
+++ b/numpy/core/src/multiarray/lowlevel_strided_loops.c.src
@@ -302,34 +302,6 @@ _swap_pair_strided_to_strided(char *dst, npy_intp dst_stride,
 }
 
 static void
-_strided_to_contig(char *dst, npy_intp dst_stride,
-                        char *src, npy_intp src_stride,
-                        npy_intp N, npy_intp NPY_UNUSED(src_itemsize),
-                        void *NPY_UNUSED(data))
-{
-    while (N > 0) {
-        memcpy(dst, src, dst_stride);
-        dst += dst_stride;
-        src += src_stride;
-        --N;
-    }
-}
-
-static void
-_contig_to_strided(char *dst, npy_intp dst_stride,
-                        char *src, npy_intp NPY_UNUSED(src_stride),
-                        npy_intp N, npy_intp src_itemsize,
-                        void *NPY_UNUSED(data))
-{
-    while (N > 0) {
-        memcpy(dst, src, src_itemsize);
-        dst += dst_stride;
-        src += src_itemsize;
-        --N;
-    }
-}
-
-static void
 _contig_to_contig(char *dst, npy_intp NPY_UNUSED(dst_stride),
                         char *src, npy_intp NPY_UNUSED(src_stride),
                         npy_intp N, npy_intp src_itemsize,
@@ -381,7 +353,7 @@ PyArray_GetStridedCopyFn(npy_intp aligned, npy_intp src_stride,
                 }
             }
 
-            return &_strided_to_contig;
+            return &_strided_to_strided;
         }
         /* general dst */
         else {
@@ -408,7 +380,7 @@ PyArray_GetStridedCopyFn(npy_intp aligned, npy_intp src_stride,
 /**end repeat**/
                 }
 
-                return &_contig_to_strided;
+                return &_strided_to_strided;
             }
             else {
                 switch (itemsize) {
@@ -445,7 +417,7 @@ PyArray_GetStridedCopyFn(npy_intp aligned, npy_intp src_stride,
                 }
             }
 
-            return &_strided_to_contig;
+            return &_strided_to_strided;
         }
         /* general dst */
         else {
@@ -462,7 +434,7 @@ PyArray_GetStridedCopyFn(npy_intp aligned, npy_intp src_stride,
 /**end repeat**/
                 }
 
-                return &_contig_to_strided;
+                return &_strided_to_strided;
             }
             /* general src */
             else {
@@ -578,7 +550,7 @@ NPY_NO_EXPORT PyArray_StridedTransferFn *
 /**end repeat1**/
                 }
 
-                return &_contig_to_strided;
+                return &_strided_to_strided;
             }
             else {
                 switch (itemsize) {
@@ -626,7 +598,7 @@ NPY_NO_EXPORT PyArray_StridedTransferFn *
                 }
             }
 
-            return &_strided_to_contig;
+            return &_strided_to_strided;
         }
         /* general dst */
         else {
@@ -643,7 +615,7 @@ NPY_NO_EXPORT PyArray_StridedTransferFn *
 /**end repeat1**/
                 }
 
-                return &_contig_to_strided;
+                return &_strided_to_strided;
             }
             /* general src */
             else {