Won't start on machine with Atheros chipset

[Scott James Remnant (Canonical)]

The version of hal shipped with warty core dumps on startup if you have a WiFi
card with the Atheros chipset (one of the more common ones in modern laptops).

The code that crashes it was stripped out again upstream due to large numbers of
problems.

[Matt Zimmerman (mdz)]

At this point in the release cycle, I think I'd prefer that we disable the code
in our existing hal, rather than upgrade hal and dbus. What is your opinion?

[Nathaniel McCallum (nmccallum)]

Its actually more than just atheros, other wifi cards have problems as well.
I'm up for either solution, though admittedly, I already have a patch to fix it
in the current hal (applies agains vanilla, don't know about debian). The only
real downside is having to support a block of code that upstream doesn't
support. From a support perspective, we may want the latest hal and dbus
(numerous other bug fixes, don't know if any of them affect us) because they are
both maturing at such a rapid rate. Your call.

[Matt Zimmerman (mdz)]

Can you attach the patch to this bug?

[Nathaniel McCallum (nmccallum)]

Created an attachment (id=27)
Patch to fix crash with some wifi cards

[Nathaniel McCallum (nmccallum)]

This patch only fixes the crash. It does *not* actually remove the problematic
wifi code.

[Nathaniel McCallum (nmccallum)]

What do we want to do?

[Matt Zimmerman (mdz)]

Give the new hal+dbus from unstable a try on Warty and see how it goes; maybe we
should sync them. I'm a bit concerned that the most recent uploads were NMUs,
though.

[Nathaniel McCallum (nmccallum)]

The NMU uploads were by Daniel Stone. I think we can trust him :). Only two
small problems as I see it. The new dbus has a build depend on jadetex 3.13-2,
but jadetex 3.13-1 is in warty. Once I installed that, dbus and hal built fine.
 Hal worked like a champ. The only other small problem is that Martin's work to
make hal run as non-root would have to be redone for the new version.

[Matt Zimmerman (mdz)]

CCing Martin since his changes will need merging

[Matt Zimmerman (mdz)]

I don't see a depends or build-depends on jadetex in either package; is it being
pulled in by something else? 3.13-1 and 3.13-2 are almost equivalent; 3.13-2
mostly just backed out 3.13-1.1.

[Nathaniel McCallum (nmccallum)]

When I installed the build depends, jadetex depended on tetex-bin and
tetex-extra, which had blockers agains that version of jadetex. I'm not really
sure why, it may just work fine (it may just be a glich in my system).

[Matt Zimmerman (mdz)]

The relevant Debian bug is: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=264319

where the submitter states that 3.13-1 works fine, but for some reason the
tetex-bin maintainer decided to depend on 3.13-2. We can either change the
dependency, or sync jadetex at the same time for simplicity (after reviewing the
diff, which should be minimal).

[Martin Pitt (pitti)]

I updated the privilege modifications for the current Debian unstable version.
The package works fine on sid, but it cannot be built on Warty since it
build-depends on python2.3-dbus (>= 0.22), but Warty only has 0.21.

Daniel, are the changes of 0.22 so crucial that we need it in Warty? Or can we
just modify the build-dep?

[Daniel Stone (daniels)]

0.22 actually changed APIs and stuff, so yah, it's non-trivial. I'd probably
recommend syncing 0.22 back if it wasn't too invasive.

[Matt Zimmerman (mdz)]

Let's just apply the patch, then

[Nathaniel McCallum (nmccallum)]

We've known all along that we needed to resync both dbus and hal (see comment
#7). What suddenly changed?

[Matt Zimmerman (mdz)]

comment #7 had a pretty explicit "maybe" in it. It's a bit late in the release
process to be breaking compatibility and such just for the sake of fixing this
bug, which we have a patch to work around.

Daniel, who made the upload to Debian, seems to be recommending against it as well

[Nathaniel McCallum (nmccallum)]

ok, np I'll merge in the patch into the current version

[Nathaniel McCallum (nmccallum)]

applied patch