vulkan/gpu_buf: avoid segfault with unusually aligned size_base

This patch prevents a segmentation fault that occurs when size_base is
not a multiple of max_transfer. If size_base is larger than the
max_transfer value but is not a multiple of it, the last iteration of
the loop will transfer too much. It needs to transfer how much is
remaining of size_base.

Note that the PL_MIN behaves identically when size_base is less than
max_transfer, because xfer will never be nonzero in that scenario.

Signed-off-by: Leo Izen <[email protected]>

src/vulkan/gpu_buf.c

@@ -372,7 +372,7 @@ void vk_buf_write(pl_gpu gpu, pl_buf buf, size_t offset,
         for (size_t xfer = 0; xfer < size_base; xfer += max_transfer) {
             vk->CmdUpdateBuffer(cmd->buf, buf_vk->mem.buf,
                                 buf_offset + xfer,
-                                PL_MIN(size_base, max_transfer),
+                                PL_MIN(size_base - xfer, max_transfer),
                                 (void *) ((uint8_t *) data + xfer));
         }