On Thu, Mar 18, 2021 at 2:33 PM Henri Verbeet hverbeet@gmail.com wrote:
On Wed, 17 Mar 2021 at 13:35, Matteo Bruni mbruni@codeweavers.com wrote:
@@ -422,6 +422,10 @@ void texture2d_read_from_framebuffer(struct wined3d_texture *texture, unsigned i unsigned int i; BYTE *mem;
- TRACE("texture %p, sub_resource_idx %u, context %p, src_location %#x, dst_location %#x.\n",
texture, sub_resource_idx, context, src_location, dst_location);- wined3d_texture_prepare_location(texture, sub_resource_idx, context, dst_location); wined3d_texture_get_memory(texture, sub_resource_idx, &data, dst_location);
We shouldn't need this (texture2d_read_from_framebuffer() is ultimately a helper for wined3d_texture_gl_load_location(), which calls wined3d_texture_gl_prepare_location() before calling any of its helpers), and in principle we don't make prepare calls in equivalent functions.
I think this helped to some degree for backbuffer ORM, at some point in the past. I need to recheck but I agree that it would be the responsibility of the caller to prepare the destination anyway. Actually, maybe this was "required" by patch 5/5.