new kernel ... / hardware watchdog

[Werner Almesberger]

Andy Green wrote:
> The hardware watchdog is the only one worth the dog biscuits if we do
> anything about this.

Agreed. A dead kernel or an unresponsive or otherwise dysfunctional GUI
should be a much more common sight than a user space that gets stuck in
limbo.

> On removing the code, it's easy to forget we only look at it to get the
> driver upstream.  I don't know why upstream folks would automatically
> hate on "controlled reboot on button hold" feature in the driver.

I guess I'm giving upstream reviewers a bit of a nasty reputation by
frequently stating that this or that stuff won't be accepted. They're
not a bunch of cynical bastards who enjoy putting others down.

But there are a lot of issues one has to pay attention to in such a big
shared project that would never be considered relevant in isolated
projects.

To give just one example, hardly anybody ever writes a new driver
completely from scratch. We all look at other code that has similar
functionality or that uses the same parts of the infrastructure. This
means that anything we put into the kernel today can become somebody
else's example tomorrow, and it is to everybody's benefit if this
example doesn't lead people astray.

Most of the time, it's us who are looking for guidance (*), and we
wouldn't like the examples we use to be confusing or full of hidden
mistakes either.

(*) Like in "everybody is a foreigner in most places on earth".

- Werner