Re: Detecting default signedness of char in ext4 (despite -funsigned-char)

[Linus Torvalds <torvalds@linux-foundation.org>]

On Wed, Jan 18, 2023 at 11:14 AM Eric Biggers <ebiggers@kernel.org> wrote:
>
> Now, we seem to have gotten the "let's break userspace, lol" version of Linus
> today, not the "SHUT THE FUCK UP, WE DO NOT BREAK USERSPACE" version of Linus

Heh.

Note that the reason I'm so laissez-faire about it is that "broken
test case" is something very different from "actually broken user
space".

I haven't actually seen anybody _report_ this as a problem, I've only
seen the generic/454 xfstest failures.

And "test failure" is simply not the same thing as "user failure".

Test failures are interesting in that they can most definitely
pinpoint the source of _potential_ user failures, but sometimes they
are just esoteric corner cases that don't happen in reality.

So the fact that we have had this bug since forever makes me suspect
absolutely nobody cares in real life. Yes, what's new is that it
happens on the same machine, but people have definitely moved ext4 USB
sticks around etc. I've most definitely done that myself, and it's not
been just between x86 machines.

Of course, it may also be that the filesystem patterns when you move a
USB stick around is very different from, say, the root filesystem
where you _don't_ necessarily tend to do it. So maybe the lack of
reports over the decades is not because people don't use xattrs with
the high bit set in the xattr names, but because it only happens in
situations that don't have that filesystem movement.

I dunno. On my system, at least, there is absolutely no sign of any
odd xattr names, according to something disgusting like

   find / -xdev -type f -print0 | xargs -0 getfattr

but who knows.

              Linus