davilla Wrote:Might try reading the thread, if you stick with the shipped 10.6.3 OSX version, it seems to work fine. It's only when updating to the 10.6.4 update that goofy things happen.
Still, we need to track this down. The new mac mini is just too expensive to *only* use for xbmc, and thus to feel that leaving the machine held back at 10.6.3 is an acceptable solution. In my case, it was bought with the intention of *also* using it for EyeTV and for iTMS-bought content (with switching between the three seamlessly via remote left as a problem for the future) *and*, being the Server edition, to be used for server-ish things.
I have a 10.6.3 and a 10.6.4 installation side-by-side on my machine just for testing this, but I don't know what to test next. Really need some guidance on what else to test for to narrow it down. The fact, for instance, that hardware accell still works using quicktime suggests a software fix is possible, that it might point to a suboptimal use of the VDADecoder API?
Also of course, a: this mac mini should be an ideal media player machine, so there'll be more and more people with this problem. And the slightly different versions of the mac mini 10.6.4 and that available for other macs are probably going to disappear with 10.6.5, at which point it's possible *everyone* will have this problem. And if it's a problem Apple needs to be told about to fix at their end, it would be best done before then, no? :-)
nb: what i just said about quicktime; that's worth testing out specifically. I have a memory of it playing videos that xbmc under dc:vda_h264 was struggling with, but I should check that more explicitly (given the patterns we've since seen on *which* files have a problem in xbmc). I'm afraid I don't know a way of definitively checking quicktime *is* using hardware accell other than watching cpu usage while it's doing so and hopefully seeing it's pretty low.