nvidia/fglrx expedited SRUs

[Bryce Harrington]

On Tue, Sep 04, 2012 at 03:15:50PM +0200, Soren Hansen wrote:
> 2012/9/4 Bryce Harrington <bryce at canonical.com>:
> > On Mon, Sep 03, 2012 at 10:37:55PM +0200, Martin Pitt wrote:
> >   3.  Invariably someone will comment on the SRU bug that they have a
> >       regression with the update, that they don't have with stock.
> >       Who knows if they do or not, or even if they're testing the
> >       right thing.  But this stops the process cold.
> 
> Can you help me understand what reason we have to believe that their
> feedback isn't indicative of real problems?

Yes, three separate reasons.

First, by this stage the driver in question has received widespread
testing with positive results.  So, any new regression feedback at this
stage is anomalous.  That doesn't mean it's wrong; obviously the person
had some sort of severe problem.  But it's weird that an issue would pop
up at this point, and we'd want strong proof that it's not just user
error or some other unrelated issue, and it just coincidently looks to
be correlated to the new version.  (An example scenario is that they have
a customized or third-party kernel installed.)

Second, the reporter in question does not provide details.  They assert
they're seeing a regression and assure us it's due to the new version.
They could well be right that there's a real problem, but at this stage
no evidence supports that it's anything but anecdotal.

Third, no one else reports the same problem.  Usually when there is a
real problem in the graphics driver, more than one person will see it.
Often there will be discussion in upstream forums or elsewhere that can
be referenced.  When it isn't, we're left with a bit of a mystery.


If any one of these three things are not true, we have something
actionable we can work on.  When we run into trouble is when all three
are false; we don't *think* there's a real problem, but it's hard work
for us to prove a negative.

At the end of the day, we're providing two different versions of the
driver because we know one may well work better for some people but not
everyone.  So, if there are some corner cases that the update doesn't
work as well as stock, well that's why we provide the update as an
opt-in and provide the stock driver in parallel.

Bryce