KS2009: Regressions

We *have* that. The problem is that users tend to have stranger hardware
and stranger configurations than anything that we can actually *test*:
this is as opposed to Ingo's randconfig does-it-boot testing, which while
otherwise excellent doesn't spot problems where drivers not needed for
booting and building kernels don't work properly. I've had at least one
regression with every one of the last four released kernels, and not one
of them was something that would have been spotted by Ingo's randconfig
testing (because that testing had wiped out all the low-hanging fruit
already).

However, all those regressions got squashed *fast* once I reported them:
even the one I didn't realise was a regression until someone else reported
it (the stuck-keys-in-X-on-SMP 2.6.31 locking bug). The key is to report
and test :)