Implementing num_secs_to_wakeup support

[Gabriel Burt]

On Tue, Aug 5, 2008 at 1:58 AM, Richard Hughes <hughsient at gmail.com> wrote:
> On Mon, 2008-08-04 at 17:51 -0500, Gabriel Burt wrote:
>> Would the best way to fix this be to modify pm-utils/functions to look
>> for an environment variable, say $PM_NUM_SECS_TO_WAKEUP, and use
>> rtcwake to do the suspend/hibernate/etc if it's set?
>
> Yes, the problem is the poor support for wakeup on different hardware.
> Some hardware wakes up at the wrong time, some just doesn't wake up, and
> the minority actually do wake up at the right time.
>
> It's a tricky problem.

Is it mostly older hardware that doesn't work, or is new hardware
still unreliable?  My newish Dell laptop worked fine w/ 20s wakeups
and an 8hr one with a modified rtcwake[1].  Maybe we should collect
info on what hardware is reliable on the HAL wiki?

Or, could we automatically detect if RTC hardware is reliable?  Is it
enough to set the alarm to something and then read it back, and if
it's the same, things are good?  That certainly seems like a good way
to detect if the hardware supports alarms over 24 hours in the future.

Gabriel

[1] http://marc.info/?l=util-linux-ng&m=121791353728402&w=2