On Sat, Jul 03, 2004 at 10:43:20PM +0200, Sven Luther wrote: > > The /dev/psaux situation seems to be a big mess. I added /dev/psaux > > upstream long ago so I wouldn't have to change my XF86Config for 2.4 > > vs 2.6. In the meantime X can have two different input devices > > specified and won't fail if one of them doesn't work but gives strange > > results when both work. And Debian has both on the default XF86Config. > > I don't think so, the current debian package 4.3.0 generated XF86Config > has psaux as corepointer, and input/mice sending coreevents. If the usb > mouse is missing, no big problem, but if the ps2 one is mising, X > refuses to start.
Aiih, crap. Okay, the situation is more complex than I thought. > > Currently the Debian kernel has another config option to have /dev/psaux > > support in the kernel but disabled by default. I really hate that hack > > and would just remove /dev/psaux from the Debian kernel as apparently > > our X packages don't need that transition-aid. > > Yep, they need. I would suggest a solution as follows : > > 1) the XF86Config file is debconf managed : We query the database, inform > the user with a low priority debconf question if it is set to psaux, > and change it for him. This means a priority normal install will > automatically make the change, so no user intervention is needed. > Problem 1 : if you want to run 2.4 in parallel you are screwed, a > solution would be to have the input/mice being the core pointer > always, and the psaux sending core events. > Problem 2 : kernel-package's scripts need to be debconfified. They > need to be that anyway though in order to break debian-installer less. > > 2) The XF86Config file is not debconf managed : We parse the file in the > post inst to check if psaux is the core pointer, and inform the user > to fix it himself. After all, he is managing the file by hand and > should know how to do this. I don't quite like either of those. I'll try to take a look at the X Code and/or talk to the X folks on whether we could fix X to autoprobe for a corepointer.
