Re: REGRESSION: "spi: add of_device_uevent_modalias support" and following "fix" breaks Macchiatobin

[Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>]

On Mon, Sep 20, 2021 at 2:56 PM Andreas Schwab <schwab@xxxxxxx> wrote:
>
> On Sep 20 2021, Russell King (Oracle) wrote:
> >
> > Module autoloading worked before.
>
> Nope.

It clearly did, Andreas. At least on Macchiatobin.

Maybe it didn't work for some other cases, but the point of
regressions is that they are things that broke that _used_ to work.

Other cases - that never worked - are not regressions if they continue
to not work.

And the thing that makes regressions special is that back when I
wasn't so strict about these things, we'd end up in endless "seesaw
situations" where somebody would fix something, it would break
something else, then that something else would break, and it would
never actually converge on anything reliable at all.

That is why the regression rule exists. It is NOT ACCEPTABLE to fix
one thing, and break another. That's not a fix at all. That's a
regression, and no amount of "but but but it fixes something else" is
valid.

So don't make completely garbage arguments. Russell reported a
regression. Denying regressions is not an option.

Yes, we have situations where even regressions don't matter - like
major security issues that simply cannot be fixed other ways, because
the regression _was_ the security hole.

(Or, actually, more commonly, the "nobody noticed" hole: a regression
is a bit like Schrödinger's cat - if nobody is around to notice it and
it doesn't actually affect any real workload, then you can treat the
regression as if it doesn't exist),

            Linus