Gerd Knorr wrote: > I'm feeling a bit nervous about putting bttv 0.8.x as-is into 2.4.x. > On the other hand the rewrite fixes a number more or less serious > problems: > > * Stability. 0.7.x has no known bugs in the capture code, but > the 0.8.x redesign seems to have fixed some hidden ones. > * Design bugs (which have some ugly side effects, when using vbi + > capture at the same time for example). > * pci/dma api. 0.8.x uses the pci_* functions everythere. In > contrast, 0.7.x doesn't even compile on some architectures > (sparc for example). > > Comments? I think it might be an idea to simply wrap the v4l2 parts of 0.8.x up in some #ifdefs, and put it in the kernel. As I have noted earlier, 0.7.x does not work on any of my machines, and from what I have seen, many other people also experience random hangs with it. I have looked long and hard for the problems, but they remain very stubornly hidden. 0.8.x seems to work just fine though, and seems to be a much better driver. I would also suggest (IMHO) removing the "sloppy" option, and let application writers fix the applications to work correctly with the formal V4L API (although it may not be that important to enforce the API if it will soon be superceded by v4l2...). -justin