On Saturday 08 July 2006 03:23, Trent Piepho wrote: > On Fri, 7 Jul 2006, Andrew de Quincey wrote: > > Combine this with the first problem, and you realise you have to: > > a) Remove the releases from dvb_unregister_frontend(). > > b) dvb_unregister_dvb_attached_frontend() is pointless. > > c) And you have make the card drivers do the detach themselves so they > > can be flexible about what they do. > > > > This is exactly what js suggested: he saw this issue long before I did. > > > > So I've committed a new patch to my repos converting all the previous > > code to handle this. > > Is there any point in using dvb_attach() for dvb-pll? It seems like every > single card uses it, Yeah - it is a bit pointless that; no problems with changing it here. > and there are still static calls to > dvb_pll_configure(), so it doesn't accomplish anything. At least not so > far. Not for long :) > The call to ops.i2c_gate_ctrl() after the tuner_ops.release() is missing > now. Does that matter? Not so much; the frontend is being shutdown as well at that point, so it doesn't really matter. The frontends should really be patched to close the i2c latch on release(). > There is a change in the way a NULL release method is handled. It used to > be that a warning message was printed if the release method was missing. > Then it was upgraded to BUG_ON if it was missing. This latest behaviour is > to silently ignore a missing release, which is a significant change in > direction. Well - the reasoning was that right now you _have_ to implement a release() method for the demodulator. However, this complicates cards that only ever have a firmware "frontend" e.g. the Nova-USB-DEC cards. Removing the release() check means cards like this can just set the various ops() calls as they want in the one piece of code, and not bother with a release if its not necessary. The original reason I added that check was so during the demodulator refactoring I could see if I had missed a release call out. I actually meant to remove it after that was done, but forgot. > Along with that, the dvb_detach() stub version is somewhat different than > the real version. Real version: > Silently handle NULL function pointer > Set function pointer to NULL when done > > Stub Version: > Crash on NULL function pointer > Doesn't set function pointer to NULL when done Yes, thats not too great. Will fix. _______________________________________________ linux-dvb@xxxxxxxxxxx http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb
