Kenneth Aafl?y wrote: > On Friday 18 March 2005 01:24, Oliver Endriss wrote: > >>Kenneth Aafl?y wrote: >> >>>Istead of having the stv0299 inittab with slightly different values all over >>>the place, what about having a default inittab in stv0299.c, and replace the >>>inittab field of stv0299_config with a patch_inittab function that is called >>>after the default have been loaded? >> >>Of course, it could be done this way, but... >> >>If you modify the shared global table you will automagically touch all >>drivers which depend on this frontend. Usually you don't have the >>hardware to test whether all drivers still work with the new settings. >> >>I don't know how others think about that. I'd like to keep the drivers >>as stable as possible. (Don't touch working drivers unless you can test >>the modifications.) So I'd rather waste a few bytes and keep changes >>local. >> >>How do others think about this issue? > > > Just to clarify what I meant above: > > I did not mean that the patch_inittab should modify the default inittab > in any way, just overwrite the registers that the adapter/frontend type > require. Maybe patch_inittab was a bad name, post_inittab would probably > be a lot clearer. As an initial change, the inittab in av7110.c will > be copyed to stv0299.c and the others diffed to this table to extract > the differences. The policy for changing the default inittab in stv0299 > would then be that all the post_inittab functions change a particular > register to the same value. > > Also I belive your statement about not touching a working driver at all > is somewhat naive, there are levels of changes that are acceptable. I understood what you meant and I think so did Oliver, but I agree with Oliver that an own complete inittab for each driver would be far easier to understand and maintain than separated patches spread over different files. But since I'm not an active developer, just my 2 cents. > If we could figure out why flexcop/skystar2 and budget-ci nearly rewrites > the inittab compared to others those it would be a lot better. > I totally agree with that. Porting the skystar2 driver to revision 2.6 would have been easier if the differences wouldn't have that large. But I don't think that separating the inittab does help here. Niklas
