i used to a good while ago

had an atheros based card for hostap

ketas, Things are improving in that area: freebsdfoundation.org/blog/how-to-u…lock-high-speed-wi-fi-on-freebsd-14

That's all about improving kernel driver performance and not user interface. But kernel wifi driver performance has been really very slow so that's not a bad place to improve things.

yes it is

anyone using that wants to use current

For the kernel driver wifi performance improvements listed there the changes in CURRENT has been merged into 14.3-RELEASE.

it somewhat improves laptop sta

but it's still being worked on

and nevermind exotic setups

to be honest, there's not much hw one could use to create ap's too

there'a good support with pcie

but there's no hw to stick pcie card into

there's ton of hw where you can stick usb into but that's massive shitshow

it does work almost

usb does add latency and has no dma but that's only part of issue

only benefit is that doorstop hw is available at doorstop price

there's also invisible stuff going on but public reports involve someone's intel wifi card in laptop is having trouble having 5ghz connection to already overheating comcast router

also important but

after upgrading from 14.2 to 14.3, and also pkg, the fish shell has stopped working for me

it dumps core, with "Bad system call (core dumped)"

immediately on trying to start

does anyone know what is going on with it?

the question is, which system call?

running it under `truss` may be enlightening

that is a very good idea but the output is long and I don't know what it means

This looks suspicious: compat11.kevent(0,0x0,0,0x0,0,0x0) ERR#78 'Function not implemented'

could it be related to freebsd 11 binary compatibillity? That's what compat11 makes me think of

yeah

iirc we smack ENOSYS with a SIGSYS by default

but we do build COMPAT_FREEBSD11 (and pretty much every other compat option) into GENERIC; are you just running stock packages/configs, or custom stuff?

I see. Does GENERIC have compat11? I've been compiling my own kernel for years, and fish has been running fine until this update

ah, sorry, I was too slow

yes, it's my own kernel config

yeah, it has all of the compat options

I find it strange that it would start needing compat11 now, since that's several years out of date

yeah, that's quite odd

ohhh

> The biggest change is that the entire shell has been rewritten in Rust.

there's your answer

flipping fantastic ...

thanks for the help

yep, sorry to be the bearer of bad news

at least I gained some knowledge :)

why is iron oxide so popular anyway

Why is rust leveraging compat11? That's gotta be a mistake somewher

Just installed fish shell via pkg -- works great.

rust is popular because it generally makes doing the dumb things harder than doing the right thing.

wavefunction: not a mistake, hysterical raisins

trying to get them to move on is a whole thing

I believe this is the one

you know it's a good one because it's been ongoing from 2021

ivy: I thought it was generally agreed upon that gnu coreutils are shit :)

polarian: Rather, most non-Linux Unix-like OS users agree that GNU coreutils are shit, as well as GNU in general.

But at least they're not the ones sabotaging Linux from the inside, that's IBM and all the products they own (Red Hat, Fedora, Gnome, Wayland, SystemD, DBUS, Avahi, and more).

Microsoft sabotages Linux too, but at least they also add some kind of value back in return.

So they at least kill Linux with respect.

anyone here runs Forgejo on FreeBSD? I'm getting some weird issues about "Command error: stat /usr/local/sbin/custom/conf/app.ini: no such file or directory"

antranigv: hah, the gitea package had that problem too

but i don't remember what fixed it

antranigv: sounds like FORGEJO_CUSTOM is broken

it should be set in the forgejo rc sript to /usr/local/etc/${name}

rtprio I remember that too. I'll check an older machine archive now.

kevans indeed. and it is set btw

odd

well that's the extent of my knowledge :-D

i do remember the port overwrote my custom decorations, but i never bothered to set up a pr for it

kevans good enough!

it helped! ok I think I know what the issue is

oh?

kevans: "hysterical raisins" - I'm keeping that in my witticism booklet.

ugh, might be time to protect my gitea

rtprio: I had to starting a couple weeks ago. And that was just for cgit. gluecode.net/blog/posts/2025/anubis-installed.html

Looking at that linked rust issue, it really seems like a lot of developers *lately* don't actually care about finishing discussions/documentation/updates.

I will always appreciate the BSDs for having at least some intention of polishing a release.

yeah, i was eyeing anubis. but i could move it behind wireguard / lan only and no one else would care

If it is for your own personal use only you can always put it behind http basic auth too which is even simpler.

that would break all my shit

but yes, it would be simpler

But... but... catgirls

Is seemingly excessive inode usage a symptom of ZFS usage?

not that i've observed

where are you seeing "excessive inode usage"?

I'm not. . . trying to help someone on EFnet.

Start with each file consuming at least one inode. An excessive number of inodes used would mean an excessive number of files. Lots of small files?

That doesnt' seem to be the case.

i didn't know people still hung out on efnet

I'm wondering if inodes are not being released when files are deleted.

CrtxReavr: well, can you ask the person you're helping where they're seeing excessive inodes (and what that means)?

df -i

The chief complaint is disappearing space.

do they have a ton of snapshots and then deleted files?

they would also take up space

depending on what happened since

so he's out of space, inodes are a symptom, not the problem

Well, a file shouldn't use more than one inode.

what about a file in a snapshot

Where I've traditionally seen inode problems is when people ran out and bought multi-TB drive, made it one volume, newfs'd it, but then the default inode density was too low, so little files would eat up all the inodes.

yeah, for ufs maybe

i haven't seen or heard of an out of inode situation in like, more than a decade

zfs list -o space how much does he have in snapshots

?

I thought that inodes were dynamically allocated on zfs. Pretty sure.

Yes. Since our cgit server is suffering network abuse stress: bpa.st/6EEQ

Our poor cgit.freebsd.org/src is giving me sporadic success with mostly 503 failures. As are most source servers on the Internet. Due to AI Scraper Crawlers endlessly hammering on them.

so it ... makes it up for df -i ?

more or less?

This would be something good to have an interpretation from someone who would know but... It appears to me that it reports based upon the currently dynamically allocated values. Which means as values change you can get wild fluctuations in results for percentages and such.

Note that I am far from knowledgeable about this code base. But I think it is dynamic and so the mapping to df values is not going to be like it is on UFS.

CrtxReavr: so ... an update?

It is no longer possible to update iocage jail if the release is no longer in the updates server?