On 26 January 2010 11:51, Stefan Dösinger stefan@codeweavers.com wrote:
+HRESULT wined3d_event_query_finish(IWineD3DQuery* iface) {
Please don't create function prototypes (or other code) by copy/paste. Also, why does this take a IWineD3DQuery pointer in the first place? You only really access wined3d_event_query fields.
/* A glFinish does not reliably wait for draws in other contexts. The caller has* to find its own way to cope with the thread switch*/
...
/* Since this is an internal call the caller should know that event Query support is faked */FIXME("(%p): Event query not supported, calling glFinish\n", This);wglFinish();return WINED3D_OK;- }
This is inconsistent. Also, note that event queries are pretty generic things, you could use them for something else than draws as well.
- if (!gl_info->supported[APPLE_FENCE] && !gl_info->supported[NV_FENCE] && gl_info->supported[ARB_SYNC])
This doesn't make sense. Of course if you just used wined3d_event_query instead of IWineD3DQuery you wouldn't even need this at all, you could just let context_alloc_event_query() fail.
Just for reference, glFinish() behaviour is a bit more subtle than you make it sound. It flushes GL commands for the current context to the GPU, and then waits for the GPU to finish execution up to that point. It won't flush command buffers for other contexts, but anything that's already been submitted to the GPU before the glFinish() will have been executed when the glFinish() returns.
I didn't look at the other patches much, but I think you want to look them over a bit before resending as well.