Bug#659378: [xen dom0] gnome-shell fails to start

[Jonathan Nieder <jrnieder@gmail.com>]

Panayiotis Karabassis wrote:

> That making software free is making software better -and not only
> technically- is obvious. But we are at the mercy of the hardware
> manufacturers, at least for now, and the nouveau developers have a
> much more difficult task to achieve that the nvidia developers.

Sorry, I think I was unclear.  I actually only meant that:

 - ordinary upstream Linux kernel, Xen, and X developers are able to
   understand the nouveau driver and not the nvidia driver, and in my
   experience presumably for this reason they refuse to even humor
   reports about the latter

 - as I mentioned, there has been recent work on getting nouveau
   working in a Xen dom0.  And I have not heard of any work in this
   direction for the nvidia driver stack.

I don't actually believe nouveau is fundamentally better for humanity
than the nvidia driver.  In a way, they are a collaborative effort ---
nvidia cannot release specs, so they provide a port of their Windows
driver which the nouveau developers can reverse-engineer to produce
a driver that uses the usual Linux interfaces.

[...]
>> There has been some important relevant work in v3.3-rc2.  Could you
>> test a 3.3-rc2 or newer kernel from kernel.org?  See [2] for
>> instructions.
>
> Thanks, but it didn't work (assuming I didn't make a mistake
> compiling the kernel).

There's only been relevant work in that version for nouveau (and other
drivers that use ttm, like the radeon and i915 drivers).  Did you test
with nouveau?  If so, could you say a little more about how it didn't
work?

>                        I am confident that this will be fixed as
> newer kernels or binary drivers become available. Will you leave
> this report open until then, in case someone else has the same
> problem?

Do you mean that you would like the Debian kernel team to keep track
of when a driver they have nothing to do with gets fixed and let you
know?  I'm sorry, but they don't have time for that, and as I
mentioned before, there is already a bug filed against the appropriate
package (bug#570365).

Hoping that clarifies a little,
Jonathan