On Thu, Jan 12, 2006 at 07:34:46PM +0100, Frans Pop wrote: > On Thursday 12 January 2006 10:06, Mattia Dongili wrote: > > no, 345537 only fixed a bug, xorg was already trying to read acpi > > events. I's actually just a matter of which process starts first: > > - acpid: xorgs opens acpid's socket and we all are happy > > Does this still work when acpid has to be temporarily stopped because you > are suspending to ram and you don't want the powerbutton event to shut > down immediately after resuming? > I tested this with my laptop recently and the system and X.org came back > beautifully, but I could not restart acpid. > > I've not yet investigated deeply though.
it's a side effect of my patch at 345537. If the socket disappears then xorg tries to open /proc/acpi/event directly and when successful steals /proc/acpi/event. There's also a different patch there that simply drops acpi events if the socket disappears (I still prefer the currently included patch as it follows the same path at startup: first try the socket and if it fails open /proc/acpi/event). Anyway you need either one as the code was buggy. -- mattia :wq!
signature.asc
Description: Digital signature