[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bug#398533: RFP: tp-smapi -- exposes some features of the ThinkPad - first packaging attempt

On Fri, 08 Dec 2006, Daniel Baumann wrote:
> > You can find the packages at
> > http://debian.die-welt.net/pool/main/tp-smapi/ - I would love to see
> > much feedback, because this is my first real packaging attemt (but
> > neither lintian nor linda do complain).
> If you're looking for a sponsor, I can do this.

A few doubts about this:

tp-smapi is either patch-only, or 2.6.19-only.  How are you packaging it?
It cannot be just built out-of-tree in 2.6.18, it will not always work as it
needs to extend the DMI handling code.

Also, please make it extremely clear in the package descriptions that the
tp-smapi HDAPS is the one people should use (instead of stock HDAPS).  Stock
HDAPS is buggy and broken.

As an extra icing on the cake, stock hdaps won't load if tp-smapi is active,
and vice-versa.  But you will get confused users wondering why one of the two
is not loading... so you need tp-smapi hdaps if you have the tp-smapi
drivers loaded.

Anyway, the reason why I never packaged it is that to get a proper thinkpad
kernel, you need a small stack of patches that are still in flux and I
didn't feel it was a good idea to package yet:

	1. tp-smapi (which will get through a BIG interface change
	   soon, when the generic battery handling starts being merged,
	   and it is not going to be merged anytime soon);

	2. ibm-acpi extended fan and thermal control (got them merged
	   into the linux-acpi test tree yesterday);

	3. block queue freeze (was not working on 2.6.18 until less than
	   one month ago, real merge work will probably start now that
	   2.6.19 is out and there is some in-tree support for what the
	   patch needs), and the hdapsd userspace daemon;

	4. The new experimental ACPI bay and dock handling that are in
	   linux-acpi test tree (deprecates lt_hotswap -- out of tree bay 
	   handling), or at the very least, ibm-acpi patches for the *60

	5. Proper support for SATA (T43, R52 and newer; 2.6.16+), AHCI (T60,
	   R60, X60, z60 and newer; 2.6.19+), fixed PCI-e and MSI support
	   (T43, R52 and newer; 2.6.16 for what the T43/R52 needs, 2.6.19+
	   for what the *60 models need)...  This means a small but scary
	   patches to 2.6.18, one or two on 2.6.19, and a pletora of them on

	6. Something else I am forgetting.

Especially because of (5) above, that is required for proper power
management and hotplugging (docking, bay), I didn't bother.  Anyone would
benefit from tp-smapi is likely to want a lot more stuff that requires
"build your own kernel" anyway.

For 2.6.19, you can actually have tp-smapi work reliably as an out-of-tree
build, so packaging it starts making more sense.

  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh

Reply to: