On Thursday 28 January 2010 23:40:25 Henri Verbeet wrote:
+static void prune_invalid_states(struct StateEntry *state_table, const struct wined3d_gl_info *gl_info) +{
- unsigned int start, last, i;
- start = STATE_TEXTURESTAGE(gl_info->limits.texture_stages, 0);
- last = STATE_TEXTURESTAGE(MAX_TEXTURES - 1,
WINED3D_HIGHEST_TEXTURE_STATE); + for (i = start; i <= last; ++i)
- {
state_table[i].representative = 0;state_table[i].apply = state_undefined;- }
Please use the number of texture blend stages reported from the fragment pipeline rather than gl_limits.texture_stages. texture_stages only works with the nvrc and ffp pipelines, but not arbfp or atifs(it works out ok by chance on most GL cards).
There's a deeper issue with how we manage texture limits, because currently the extension loading code picks limits.texture_stages, already anticipating which pipeline we'll choose later. We should handle this in the way shader constants are handled: Rename limits.texture_stages to limits.max_nv_general_combiners, make only the nvrc and nvts pipelines read this extension specific value, and use the pipeline reported d3d texture blend stages in fragment pipeline independent code. We're also (still) mixing up blend stages with fixed function vertex coordinates.
Regarding masking out unsupported / invalid states, Ivan once suggested the idea to make the pipeline template's needed extension field more powerful, e.g. via a callback function. That way the pipeline would have more control over which state handlers it registers. I am fine with a pipeline independent function, I guess it will work just fine for our needs, just digging out the old idea.