On Sat, Oct 03, 2009 at 01:49:42PM +0200, Manuel Lauss wrote: > On Sat, Oct 3, 2009 at 12:22 PM, Wolfram Sang <w.sang@xxxxxxxxxxxxxx> wrote: > >> > Yeah, I saw that you want to remove it, still I don't know why :) Is it feature > >> > incomplete and updating is impossible? Is the concept outdated? Could you > >> > enlighten me on that? > >> > >> I started out with the intention to fix its styling issues, add carddetect irq > >> support, etc. In the end it was easier to write a quick-and-dirty standalone > >> full-features socket driver for the DB1200 and extend it to support the > >> other DB/PB boards. While I was at it I modified my driver for the xxs1500, > >> that's all. > > > > Okay, that explains. > > > >> > >> The only *technical* reason I have is a personal dislike for how the current > >> one works: it forces every conceivable board to add dozens of cpp macros > >> for mem/io ranges and gets registered by board-independent code. > >> Hardly convincing, I know. > > > > Well, you have the (to me) pretty convincing technical argument that your > > drivers provide more features and less crashes which is a clear benefit for > > users. If we remove the generic au1000-part, then it might even be in the same > > amount in LoC. Okay, we lose a bit of maintainability if a bug is found in a > > section which was shared among the former users of generic, as it has to be > > updated for each of the three drivers, but well... Are there any plans to > > convert pb1x00 as well? > > The new db1xxx_ss.c already supports all boards pb1x00 is supposed to, > except for the PB1000 (the very first Alchemy devboard), which has a > rather awkward carddetect irq scheme, so I kept the au1000_pb1x00.c > for it. Unfortunately I don't have this board to test on, and *if* there are > any linux users with this board, they choose to remain silent (the driver > hasn't built for it in years, so go figure). I'd rather get rid of > PB1000 support > altogether... > > > > Maybe I find time to look a bit more into it, but I can't test anything, of > > course, so the more additional comments/test-reports the better. > > Thanks. As I mentioned, the db1xxx_ss part works on my Db1200/Db1300 > boards; I don't have any others to test on. Deending on the urgency you assign to these patches I can keep them in my queue for 2.6.33 and push them upstream for linux-next. Ralf
