Installing drivers on USB sticks

[Bryce Harrington]

On Tue, Jan 04, 2011 at 02:08:28PM -0800, Brian Murray wrote:
> I've recently discovered hundreds of bug reports regarding people trying
> to install drivers (bcmwl, fglrx, nvidia) on their Live USB version of
> Ubuntu and it failing[1].  Part of the large volume is due to the fact
> that the bug reports are automatically reported by apport as package
> installation failures.  When looking at these bug reports I had a couple
> of questions.

Yeah this is something we've been studying from the X.org side as well.
This is by far the most common X bug reported against natty so far.
 
> Is installing drivers on a Live USB version of Ubuntu something that is
> supported?

I experimented a bunch with it and fglrx before the holidays, and simply
could not get it to work.  It needs to update the initrd for it to work,
but that file lives outside the persistent file system.

So, the way things are designed with USB LiveCD environments, no it is
not supported currently.

People have reported success at installing drivers when they don't
require initrd updates (e.g. old versions of -nvidia).

> It seems to me that there are a lot of people who think it should be and
> I believe it would a useful way for people to test the development
> release of Ubuntu and the latest drivers.  If it isn't though we should
> stop the automatic reporting of this classification of bug reports[2].

Yes, it is clear people want to test Ubuntu this way.  It would be great
if we could enable it.  Although, given the differences in how the USB
live environment is laid out compared with a normal installation, it
might require a good bit of debugging itself(!)

For instance, it's possible with usb-creator to allocate all space to
the persistent file, which causes the new initrd to be written to begin
with.  And other problems...

Anyway, I agree we ought to either stop the automatic reporting, or make
the reports auto-dupe.

Even better would be to get Jockey to not offer to install these drivers
in the first place.  (E.g. bug #685017)  When we feel it is a feature we
definitely do feel comfortable supporting, then we could re-enable it.

> There are also some bug reports[3] that have empty
> DpkgTerminalLog.txt[4] attachments.
> 
> How did this happen and can it be fixed?
> 
> Can these be marked as duplicates of bug 557023? (Provided they are on
> live media and about installing drivers of course.)

Yes, especially for -nvidia and -fglrx feel free to dupe them up; as it
is the reports are just causing noise for us.

Bryce