Bcachefs Fixes Pull Once Again Frustrates Linus Torvalds - Two Choices Offered

[mobadboy]

So it´s a project for the ELITE you are not entitled to be here .

I better not spew out all the bull thats just going trough my head.

leader start project
leader set rules for contributors and maintainers
contributions and maintainers agree to rules
contributor Kent not care about small details, rules
contributor Kent break build for very important architecture
many people say wtf lol
leader says Kent stop breaking my builds (aka go away)
problem solved

OR

leader start project
leader decide hey, who cares right. carpe diem
different contributors do different things because one care about ia64 (contributor A), the other not (contributor B)
contributor B breaks build for ia64
contributor B say hey why do i care, nobody uses ia64 anyway
contributor A say my employer does
leader not care, says fuck u both
impasse achieved
product rots

[TheMightyBuzzard]

So it´s a project for the ELITE you are not entitled to be here .

I better not spew out all the bull thats just going trough my head.

Probably a good idea since it makes you sound like an entitled child. It's Linus' project and he has every right to run it how he sees fit. It doesn't make him any more elitist than you running company off when you're tired and want to go to bed. Your house, his kernel.

[You-]

So it´s a project for the ELITE you are not entitled to be here .

I better not spew out all the bull thats just going trough my head.

Its opensource. Everyone can pull their own tree. You can even have your own tree which if it becomes popular gives you the power to decide what is in it.

Many Arm vendors also have their own trees which often differentiate very significantly from mainline - and everyone hates them for it. The raspberry pi foundation also have their own pseudo tree which has a lot of code that has not yet been upstreamed.

In general the community have liked how Linus does it so that has been the main tree. In the past before he discontinued it I think it was Andrew Morton who had a separate fairly popular tree. He discontinued it in favour of Linus' tree.

[Quackdoc]

As one of the users that Kent talks about, Bcachefs being in or out of tree doesn't have that much of a bearing on whether we'll use the file system or not. I've used a certain out of tree file system, I don't believe I gotta mention its name, for around 8 years now. I just pick distributions with installers that make it easy to use that out of tree module, like I did when I used graphics cards that required out of tree modules and like I'll do going forward for Bcachefs if it gets kicked from mainline.

It really doesn't matter if something is mainline or not. What really matters is both how well something is supported by your favorite distribution and how easy or hard the developer makes it to use on different distributions.

it was the major deciding factor for me personally. Im not going to recompile a kernel every update lol. kent has stated multiple times to not use DKMS too, if DKMS had better support I wouldn't have an issue. I can DKMS any day, make my own if really needed.

[fitzie]

Kent is too chaotic and emotional. Not that no chaos is good, he's pushing for a lot of good things, like the the inode cache changes happening due to bcachefs using rhashtable https://lore.kernel.org/linux-fsdeve...6b3d4999aa52b9

But that doesn't excuse his basic breaking of the rules. I won't rehash all the old ones, but it's really just down to two that kent seems hellbent on ignoring:
1) posting his patches to bcachefs mailing list
2) not sending his patches to bake in linux-next

These are what linus are talking about, but I suspect that he's also talking about his unnecessarily combative behavior in general and other rule and norm breaking, like going around maintainers, etc.. When linus tells kent just to maintain an experimental tree, this is something I don't understand why kent doesn't adopt. when he had critical bugs in stable bcachefs, he refused to put out a tree for users to get the fix, but instead complained to stable tree maintainers for not pushing out a release right away for him, but there would always be a delay between the fix and stable/upstream, so there's a good reason for kent to maintain these tress. Kent should just formally promote his trees for bcachefs users, and follow the rules for upstream/stable submission.


The real question is why does kent not get the hint? Well, sadly, because there are no repercussions. As sad as it is that linus has to give kent an ultimatum, it's even worse that this isn't enough to get kent to change. Linus will have to not accept kent's patches for a kernel or two for kent to get the message. I'm actually upset that linus didn't start now.

I supported bcachefs on patreon for many years, and I still believe it has a bright future, but i've been outspoken here against kent's behavior because it's obviously detremental to his project, and it's entirely juvenile.

[fitzie]

Michael l actually the drama isn't on lkml but on fsdevel https://lore.kernel.org/linux-fsdeve...ernel.org/T/#t

some choice quotes from kent:

If what you want is patches appearing on the list, I'm not unwilling to
make that change.

I take issue, and indeed even dig my heels in, when the only people
asking for that are _only_ yelling about that and aren't involved
otherwise.

Let that sink in.

perhaps you and the community don't know quite as much as you
think you do about the realities of what it takes to ship a working
filesystem.

If you're so convinced you know best, I invite you to start writing your
own filesystem. Go for it.

[Developer12]

As one of the users that Kent talks about, Bcachefs being in or out of tree doesn't have that much of a bearing on whether we'll use the file system or not. I've used a certain out of tree file system, I don't believe I gotta mention its name, for around 8 years now. I just pick distributions with installers that make it easy to use that out of tree module, like I did when I used graphics cards that required out of tree modules and like I'll do going forward for Bcachefs if it gets kicked from mainline.

It really doesn't matter if something is mainline or not. What really matters is both how well something is supported by your favorite distribution and how easy or hard the developer makes it to use on different distributions.

I'd kinda like my filesystem to be tested. Kent seems to do nothing but throw fixes at the wall and ship them, hoping that eventually the bugs will be solved because he's hopefully fixing more bugs than he's introducing, and to hell with what happens to people's data along the way. No thanks.

[ksec]

if he is going to continue to be a petulant child and a liability i don't know why he wouldn't remove it from the project. i don't see linus as being unreasonable at all. is he expected to just let kent disrupt the entire project indefinitely? how exactly is linus being an asshole here?

Not that I disagree with you in the slightest. I am somewhat surprised by this question in 2024. I thought we are so bad in the past 10-12 years that the word toxic positivity was born.

[Developer12]

I don't think it's unwarranted for linux to pull the filesystem from mainline. Kent is clearly trying to get fixes merged that are untested, and they're breaking things for other mainline contributors (like people who support big endian architectures). It might even knock Kent's attitude down a peg or two.

I don't think it's reached that point yet, but I also haven't seen the full total of the breakage Kent's bad patches have caused for other contributors.

The alternative would be for linus to start enforcing some structure, like refusing to merge fixes until the next merge window if Kent is late. Would that actually bring Kent in line from an attitude standpoint though? Not so sure.

[curfew]

Wow another drama, actually they are both assholes that think they know better, the one that does not get his patches reviewed and the guy that is stubborn and threatens to remove the hole project is no better.

Linus already has enough "polite" and functional filesystems in his kernel, so he can boot Kent off for being naughty and nobody would miss him. Kent, however, can also just take his toy and go home and write his own kernel, because it sure seems like he knows more about kerneling than anyone currently alive on earth.