reported by Seblu(seblu@seblu.net) in the maillist,
http://marc.info/?l=linux-acpi&m=128855015826728&w=2

When starting up my laptop my AC adapter dectection was not done
correctly in sysfs, but correclty via procfs.

At first sight this was weird because acpitool calls ask my power
adapter is online, but my kde power manager dected adapter off-line.

$acpitool
  Battery #1     : charged, 100.0%, 00:00:00
  AC adapter     : on-line
  Thermal info   : <not available>

acpitool bring informations from /proc fs

$ cat /proc/acpi/ac_adapter/AC/state
state:                   on-line

So my system see correctly AC plugging state, so why kde not?

After some tracing, it seems that : kde power manager use powerdevi,
which use solid, which use hal, which use sysfs.

$ lshal -u /org/freedesktop/Hal/devices/computer_power_supply_ac_adapter_AC
udi = '/org/freedesktop/Hal/devices/computer_power_supply_ac_adapter_AC'
  ac_adapter.present = false  (bool)
  info.capabilities = {'ac_adapter'} (string list)
  info.category = 'ac_adapter'  (string)
  info.parent = '/org/freedesktop/Hal/devices/computer'  (string)
  info.product = 'Generic AC Adapter Device'  (string)
  info.subsystem = 'power_supply'  (string)
  info.udi = '/org/freedesktop/Hal/devices/computer_power_supply_ac_adapter_AC'
 (string)
  linux.hotplug_type = 2  (0x2)  (int)
  linux.subsystem = 'power_supply'  (string)
  linux.sysfs_path =
'/sys/devices/LNXSYSTM:00/device:00/ACPI0003:00/power_supply/AC'
(string)

Got it, hal says adapter is offline...

cat /sys/devices/LNXSYSTM:00/device:00/ACPI0003:00/power_supply/AC/online
255
watching dmesg show error about acpi ac adapter reading... (in pj)

The most strange for me, is, if i call acpitool or cat
/proc/acpi/ac_adapter/AC/state, value in
/sys/devices/LNXSYSTM:00/device:00/ACPI0003:00/power_supply/AC/online
the answer goes right become 1. And so everything become ok. if i
unplug and plug again the power cord, online state is reported
correctly. This issue is only when the pc start with power cord
plugged.

I believe, issue come from a race condition, because if in my init
script i access to /proc acpi file, and after hal start, status is
reported correctly.

My current workaround is to load ac module in initrd, because, distro
startup run udevadm trigger after and when it's done before, behaviour
was like i said before.

my test was done on a dell laptop e6410 under linux-next-2010-10-29
and 2.6.35.8. (2.6.36 has big drm-intel regression on my laptop and
become unusable).

$uname -a
Linux rwolf 2.6.36-next-20101029 #1 SMP PREEMPT Sun Oct 31 17:34:37
CET 2010 x86_64 Intel(R) Core(TM) i7 CPU M 620 @ 2.67GHz GenuineIntel
GNU/Linux