Kristian H?gsberg <hoegsberg at gmail.com> writes: > I'm OK with either approach. It does seem like cleaning up the DRI > driver interface is orthogonal to enabling the __DRIimage based > getBuffer callout though.
We should probably not merge what we don't want to maintain though; let's decide over lunch. I don't think it matters very much to the current code, it'll only bug us in small ways in the future. > I think that's fine. I was going to say that if we expect the > requested and the returned set of buffers to differ, we might as well > just memset the struct and let non-NULL images indicate returned > images. But in case of a driver with a newer interface that extends > the struct (stereoscopic buffers), the loader can't memset the entire > struct (it only knows the smaller, previous version), and the driver > will think the non-NULL garbage fields are valid images. So the > image_mask makes sense. That's what I was thinking. -- keith.packard at intel.com -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 827 bytes Desc: not available URL: <http://lists.freedesktop.org/archives/dri-devel/attachments/20131106/035a1826/attachment.pgp>
