Re: slowly decommission bugzilla?

[Linus Torvalds <torvalds@linuxfoundation.org>]

On Sun, 1 Mar 2026 at 07:35, Laurent Pinchart
<laurent.pinchart@ideasonboard.com> wrote:
>
> Do we really have to use non-deterministic tools that will inevitably
> produce correct-looking but otherwise useless backtraces from time to
> time, confusing developers and wasting time for everybody, when we can
> instead easily develop tools that will work in a deterministic fashion ?

Please do develop that, since it's so easy. We haven't done it in 35
years, but I'm sure you can whip something up in an afternoon.

.. and in the meantime, how about we not waste real humans' time
having *them* be very frustrated and generating non-deterministic
guesses - and inevitably incorrect analyses - of the reason for bugs.

Christ people - have you even *looked* at bug discussions? They are
*full* of those "non-deterministic and correct-looking but wrong"
guesses.

And that's the _good_ case, when you have people actively following up
on oopses.

Yes, there are real humans that are trying to use informed guesses as
way to a solution. Sometimes that guess is correct - and with decades
of experience that percentage certainly goes up - but quite often it's
a "maybe you can try XYZ".

What the hell is wrong with people? This whole "AI can be very wrong"
is obviously true.

But it's irrelevant, and equally misleading as the Sam Altmans of the
world that lie and exaggerate through their teeth to make a buck (but
when you do it for money, it's called "marketing" and is obviously ok.
Sigh).

Determinism is fine. But the world isn't entirely deterministic, and
it looks particularly non-deterministic when there just isn't enough
information.

A "good guess" is sometimes the best you can do. And there's nothing
wrong with trying tooling that does exactly that.

Should those guesses be *marked* as such? Hell yes.

But this crazy "it's AI, so it's bad" needs to stop.

               Linus