Linus Torvalds Isn't Happy With Some Of The Bcachefs Code For Linux 6.9

[sdack]

Love the white knights and social justice warriors here who make a drama for not wanting a drama. But more so do I love how Linus puts his foot down, as he should. He does not care whether his opinions and views cause a drama and only he is professional. Others are still only trying to be professional.

[ckafi]

I can respect Linus' objections about the interfaces, but I don't see what's so insane about the math implementations. The computations are entirely correct, they're neatly optimized, and there's even a PDF linked with the derivations for each formula. What more do you want? Is it just because it's 128-bit math? Also, no, MAD is not a valid substitution, because there's no way (at least to my knowledge) to increment it on-line.

[mxan]

"senseless...horrendous...disgusting...disgusting. ..nonsensical...insane...idiotic...garbage...point less...garbage...horrendous...stupid"

Maybe he’s been spending time with Emanuele Bassi

[Old Grouch]

Overstreet submitted a pull request. Torvalds reviewed the request and demurred, giving reasons, It's now a technical discussion, and Overstreet is free to make changes and submit another request. A request is not a demand, or order, even if some people appear to confuse the terms.

It doesn't need to be dramatised or personalised. This is not a Netflix series.

[varikonniemi]

Already V2 is submitted that does not promote these interfaces outside bcachefs. Seems like the only drama regarding this "incident" happened on phoronix

[toves]

I guess in Finnish there is only one word for 'spade.' Apparently Linus would not have much use for a shovel in any case.
I appreciated the point about the calculation of median absolute deviation cf standard deviation.
In my limited experience developers commonly have very limited mathematical background and no statistics beyond mean and variance.

[rrveex]

I agree that this attitude isn't the sole, or even primary cause of burnout - just that it doesn't help.

You're well within your right to disagree, of course, but IMO...

Pussyfooting around: "Your code has a lot of good things about it, great job! What would make it even better, is if - and I'm not saying you have to do it this way, please don't think of this as a critique - you looked at possibly optimizing the naming of your functions so that they flow a bit more consistently with other similar functions and enable more effective review..."

Direct, un-softened blows: "I couldn't read your code because the names don't follow established kernel logic. Rejected until that is fixed."

Needless drama: "Your idiotic code is so hard to read I could never even begin to include it, your function names are moronic, if you make it less disgusting then I can even look at the rest without puking"

I think only the first two are "professional", and only the last two are "helpful"...so I don't see anything wrong with shooting for the intersection of professional *and* helpful.

I agree with most. What you suggest, is Linus translating what's in his head (v3) to v2 so that it's easy for Kent.

I argue Linus it's higher up the importance hierarchy, so why put the work on him and not let Kent read v3 and translate to whatever he likes?

It's just a matter of effort. "calm down; re-think; emotions don't matter; re-re-think".

[Mathias]

[...]on the other hand, he turns every thing into a battle. for example, it's not good enough that he's got subvolume added to statx call, he wanted it added to 6.9 at the last minute.

You mean this?:

Can we try to make 6.9? Need to know what to put in the manpage, and I've got userspace tooling that wants to use it.

Hm, I understand that you want to see this done but I think it's just too tight. If this would've been Acked last week then we could've done it the second week of the merge window. So my reaction is to wait for v6.10. What do others think?

Christian wrote this 2 days ago and Kent didn't push back. Did I miss any of the "battle"?

You could have called the "split it up" discussion "a war" if you want:

[...] I would prefer it to be split up in this manner for [this] reason.

[disagrees] For something this small, seeing how it's used is relevant context

It needs to also be split up because fs/ and fs/bcachefs are maintained differently. [...] and for people who are cherry-picking/backporting accordingly, having to chop out part of a patch would be unpleasant.

It's a new feature, [...] this should never get backported. [...]and I'm submitting it to the VFS maintainer, so if it's fine with him it's fine with me.

I see no reason to split this up. Especially given how small the patch is.

But all I see a discussion about how it should be done...

[pabloski]

This is not a Netflix series.

Luckily for us! Otherwise Kent would have been an alien from Vega, Linus a bug wearing glasses and Linux a transgender grimoire.

[oleid]

Proof that therapy doesn't work.

Your post?