>Nowadays, people working on XFS seem to spend most of their time on distro kernel backports and dealing with AI-generated corner case bug reports that aren't user reports.

Could someone more familiar elaborate on how AIs are submitting filesystem bug reports?

It's not AI. It's a sophisticated fuzzer called syzkaller. Google runs it as a bot, called syzbot, and for every crash or corruption it finds, it tries to bisect and send automated email to the correct maintainer and to the person whose patch introduced the bug. The email contains a reproducer. Example: https://lkml.org/lkml/2023/6/20/1355, which happens to involve XFS. Do read the first reply, it explains why XFS people are unhappy about this particular "finding".

See also this LWN article about what's supposedly wrong with syzbot: https://lwn.net/Articles/917762/

That’s pretty obnoxious. I think the XFS guy was being reasonable and the Google person comes off as a whiny jerk.

If some guy at my company reported something in error, wasting my time chasing butterflies, then turned around and made his defect an uncompensated assignment for me, that conversation isn’t gonna go well.

It’s too bad too, as the overall goal of what’s being done is a worthy one.

Fascinating comment thread.

We need more humans.

I wonder how RHEL will handle this in the future.

Fedora uses btrfs by default these days, so that's likely what's coming for RHEL.

RHEL had BTRFS as a technology preview in 8 but removed it in 9. So I don't think they were that vested in it.

Not great.