1) this is clearly wrong, as Xkb can allow you to move the core pointer in other
ways without a mouse.
2) with hot-plug input devices, it is perfectly reasonable for the number to be
zero at various times (particulary on startup).
3) The damn log file ends up filling a log based file systems with junk files,
as the X server cycles again, and again, and again.
There is a meta question as to whether a display manager should keep trying or not.
One might argue this could be deferred to 7.3, but please make sure it gets
moved to 7.3's release tracker bug.
this has already been fixed in input-hotplug, where we use a virtual core
keyboard and pointer. you can't just allow things to go ahead without one,
because of the number of unconditional dereferences of inputInfo.pointer.
When was this fixed?
I ask, as we're running pretty recent bits on the OLPC system....
try nuking the checkCoreInputDevices call in hw/xfree86/common/xf86Config.c,
maybe? i know for sure that it works in kdrive, because half the time i keep
forgetting '-mouse tslib' ...
also, half the time, my grammar is always poor.
fixed in ba9f5138fc32a7a7b97bcf941bc92751b7c6c2c0 -- put Option
"AllowEmptyInput" in server flags