Do all the things like ++ or -- rants, post your own rants, comment on others' rants and build your customized dev avatarSign Up
From the creators of devRant, Pipeless lets you power real-time personalized recommendations and activity feeds using a simple APILearn More
Search - "io_uring"
Time for a REAL fucking rant.
io_uring manpages say you can set the CAP_SYS_NICE capability to allow SQPOLL to work. You can't, you still get an operation not permitted errno result.
Why? I checked, it says 5.10 mainline is required. Pretty sure I just manually downloaded and installed the Deb's myself. uname reports that I am at 5.10. So what gives?
Maintainer submitted a patch because they fucked up and made the *actual* capability check look for what's basically root permissions (CAP_SYS_ADMIN... c'mon...) and is now trying to rectify a glaring security shortcoming.
Patch hasn't been accepted or even addressed yet but they already updated the manpages with the estimated mainline kernel release as if it had made it into the release candidate. Manpages have made it into latest debs but the actual change has not.
Where the fuck is the Linus Torvalds that would ream the fuck out of shitty developers doing shitty things? The political correctness climate has discouraged such criticism now and the result... this. This fucking mess, where people are allowed to cut corners and get away with it because it would hurt their feelings when faced with pressure.
I'm not just guessing either. The maintainer has already said some of the "tone" of criticisms hurt his feelings. Yes, sorry, but when you claim 90% speedup over a typical epoll application using your new magical set of syscalls, and nobody can even get 1-2% speedup on a similar machine, people are going to be fucking skeptical. Then when you lower it to 60% because you originally omitted a bunch of SECURITY RELATED AND CORRECTNESS CHECKING CODE, we're going to call you the fuck out for fudging numbers.
Trying to maintain the equivalent of academic integrity within the computer science field is an exercise of insanity. You'd be fired and shunned from publishing in journals if you pulled that shit in ANY OTHER FUCKING FIELD, but because the CS scene is all about jerking each other off at every corner because the mean people keep saying mean things on Twitter and it hurts your feelings therefore we're all allowed to contribute subpar work and be protected from criticisms when others realize it's subpar.
These aren't mistakes anymore, it's clear you're just trying to farm clout at Facebook - maybe even FOR Facebook.
Fuck you. Do it right, the first time. Sick of shitty code being OK all of a sudden.2
Writing coroutines for io_uring. Submissions aren't working (they go through but the operation never seems to complete). Just fixed a few segfaults because of error cases not destroying coroutine handles correctly. Been on this for two days, finding more and more things I need to do that unrelated to the task at hand. Don't have issue tracking set up yet because haven't had time. Mountain of things that need to get working just for a demo is only growing. Layered maps of data structures and code flow are in my head as I'm trying to mentally debug some of this. I'm focused, completely dead to the outside world.
Then I feel a small scratch on my cheek. Three hours of mental mapping and a deep stack of thought, vanishes into thin air in a single moment.
The trade-off is worth it though.10
io_uring hardly works and yet was accepted into the mainline Linux Kernel.
Has been two weeks of headaches.2