I see no mention of using GPE, why not use that? It should provide you with the majority of what you want.

As fanoush said, some things are not handled in the kernel, however if you leave the initfs there (and running) battery charging still works fine.

Looks like some recent rootfs's here: http://www.kernelconcepts.de/~fuchs/.../experimental/

The issues were as follows (the last time I ran the image, which was last year). This is basically copied from my post here: http://www.oesf.org/forums/lofiversi...p?t19354.html:

* Backlight control doesn't work very well - it works (writing and reading from /sys entry) for a second or two, however then the nokia stuff grabs control back and changes the backlight up to full brightness again. To fix this, you'd need to send dbus commands to change the backlight. The format has been documented afaik, so it just requires some reasonably simple patches to minilite (these may have been implemented by now).

* Sound doesn't work - This ought to be hackable, the dsp needs to be initialised, looking at the contents of the maemo image, this should be reasonably easy - just a case of copying scripts and the .bin file across.

* Power management - if anything keeps a clock running, the power management won't kick in (dyntick). Therefore you can't have the clock open as an applet on the gpe taskbar. I imagine other apps will also cause this kind of problem. Also note that maemo probably has far better control of the peripheral devices, so it's power management may be better (though that stuff may run from the initfs anyway so would work with a gpe-image, I'm not sure.)

* No battery power reporting - battery reporting is handled by dbus again. I think someone's decoded the readings, however it doesn't have very fine granularity which is a shame.

Also note Florian's comments in this ml post (it's rather old now, I'm not sure how things have changed): http://maemo.org/pipermail/maemo-dev...st/000960.html



Simon