nimaje/oxbar: thank you, I had the exact same question. nice to see its answered when I look for it, didnt even have to scroll back :-)

I guess I will try 14.1 for now

good morning! I trying to get my macbook air from 2012 with swedish keyboard layout to work.. at the moment I using this configuration pastebin.com/n7K5mQJ5 in my keyboard.conf under (/usr/local/etc/X11/xorg.conf.d/), but my FN doesnt work :( so neither brigthness or volume up/down/mute works.. I hope someone can point me in right direction to get it to work, One strange thing is when I remove 'us'

from XkbLayout and I starting X I dont get swedish it still using us keyboard layout, but when I adding it back it´s swedish keyboard..

nwe: interesting problem. it appears linux has some `hid_apple` kernel module to deal with apple keyboards. i fail at finding something similar for us.

|cos|: yes, as you mention with linux they are using hid_apple and there it´s working, but now I dont want to run linux on it ^^ I have run linux for severral years, and like freebsd more in that sense it taking care of everything not like linux everything is picked from everywhere and it glue it together.. BSD is a whole OS :)

tsoome: I reduced the panic to this commit 2eeb95ccf39ac5d79326c4482112b6b7dd5fc8b2 - reverting it fixes the problem

Hey all. I have the following ever since I started to use VNET jails, and I never really bothered to look at it. But now I'm moving some more mission critical things to this server and setup the daily root mails and get annoyed by it. can someone help please?

the ip (192.168.10.12) is configured on vlan6, which is configured on one of the physical interfaces of the server

02:ff:60:31:10:84 is the mac address of the vnet intefrace of a jail, 00:25:90:c0:63:b8 is the mac address of the physical nic + the vlan configured on the host

the jail does not have this ip configured, that's the first thing I checked

mvanbaak: i'm no network specialist. but to me the message looks like the opposite of what you say is actually the case

mvanbaak: which doesn't mean i don't trust what you're saying

ifconfig output: dpaste.org/7RCQC

jail config of the jail with the mac address in the log: dpaste.org/zgLtt

SKull: believe me, I have thought the same way for months

ifconfig inside that jail: dpaste.org/0M1CK

mvanbaak: if you have an physical interface for that ip, then why do you give the jail an epair instead of that physical interface?

the jail has a different ip. it has 192.168.10.249

the 10.12 ip is for the host

yeah, I misunderstood your problem description at first

recap: the ip 192.168.10.12 is configured on vnet6 on the host, which is configured on one of the physical interfaces of the host. This is also what I want. This way I can reach the host on that ip.

why do igb0 vlan6 and vlan100 have the same mac address?

There's a jail called 'ssh' with the ip 192.168.10.249 (using VNET) and this is also what I want

erm, good question. They are cloned interfaces

ifconfig_igb0="up"

cloned_interfaces="vlan6 vlan100 bridge0 bridge1"

o0x1eef it would be best if you can mail the details to kib@ - so he can check what is going on there...

tsoome: I reported a bug here bugs.freebsd.org/bugzilla/show_bug.cgi?id=285976 - should I cc him there?

yes, because he is the author and the commit has MFC flag - if there is indeed problem, the MFC should be postponed till issues are fixed

+1

I can drop the link for him now too.

Thanks!

did it:)

So, if someone has an idea, please let me know

mvanbaak: do you currently need vlan6 and vlan100? if not, try removing them and see if the problem continues to exist

I do :( without vlans there's no traffic to the machine

reading the handbook again, it looks like setting up vlans no longer works with `cloned_interfaces` in rc.conf, but with `vlans_<interface>` and `create_args_<vlan>` so I will update that later and reboot to see if that fixes it

Is something up with the pkg latest release? It seems packages that should be there are missing, like e.g. go124

nwe: you already knew about the linux kernel module? it was news to me, but the source code i would start looking at to see how it actually works. maybe those apple usb vendor and product id:s are possible to search for. i'm not sufficiently motivated to dig further. my macbook doesn't even have sound.

seems like there is/was a bug in -CURRENT, that resulted in a fail to build the pkg repo(?) lists.freebsd.org/archives/freebsd-ports/2025-April/007557.html and if you look at the builds themselve they have some build failures which result in many skipped packages, e.g. go* and webkit-gtk failed to build

mvanbaak: do not put IP addresses on bridge member interfaces. the IP addresses should be configured on the bridge.

hum, so vlan6 should not have an ip, it should be added to bridge0 as is done already, and the ip address should be on bridge0? is that the solution?

yes, correct

i can't guarantee this is actually the cause of your arp message but it's possible

oh wow, TIL

I will update my configuration later today and try. thanks a lot

nimaje: Ah, thanks!

|cos|: I understand which model is your macbook?

Hello people. This snippet <godbolt.org/z/bqeP7hqjb> compiles fine on godbolt; on FreeBSD-14.2-RELEASE (same clang version, 18), it just fails inexplicably.

now that I think about, godbolt might not be using libc++ by default.. *ponders*

so, I have a really strange issue right now

I'm trying to set up syncthing

and I followed the guide on the wiki, minus opening the ports because I don't even have pf running currently

and whenever I start the syncthing daemon, my network goes down

and it only comes up once i stop syncthing, and redo dhcp

can't find anything in any logs, including dmesg, ifconfig output remains the same

xcvb: i don't have a system with llvm 18 to test, but fwiw this compiles fine for me on 15.0 using llvm 19.1.7

I'm completely stumped on this one. can't find anyone having the same problem. I am on CURRENT though, could be some strange bug I suppose\

xcvb: with clang 19 it compiles fine if you explicitly request -std=c++20

what error do you get with 18?

I guess I can try a different network connection to see if it's a driver issue

mtll: i had syncthing running on 14.2 until fairly recently

I try to build py-psycopg2 from ports, but it seems that somewhere along the build process, py311-wheel044 gets installed which then later clashes with py311-wheel.

here is a build log from a fresh jail: pastebin.com/pgvrDnaB

Anyone got a clue on how to get networking to work on bhyve ? Obviously it’s not easy cause there are so many how to’s and troubleshooting guides.. I thought you just create a switch then bind it to your interface like wlan0 and it should work

I been at it trying different things since yesterday.

Setup a bridge and Nat WiFi

WiFi is special and needs complicated magic to bridge

[tj]: there a guide I can follow ?

i've only done that with wired ethernet so far, and that seemed to work okay

Bridging is bridging. Same process for any interface (providing it can be successfully bridged.)

Here are the rc.conf options I used: ifconfig_igb1="up" cloned_interfaces="bridge0" ifconfig_bridge0="addm igb1"

then, when creating the VM, i've used "virtio-net,tap0", and then added the tap interface to the bridge: "ifconfig bridge0 addm tap0"

after that, my bhyve vm was up and running

Grabunhold_: Yep. It's pretty straight forward.

i've written a small python daemon to manage it all: github.com/avollmerhaus/Anlasser

it's mostly a stupid hack, but i'm using it for my hobby projects. i should port it to async python when i have the time.

ek: that's not true on the protocol level, bridging wireless requires the 4-address packet format (WDS) - that's why wifibox does routing instead

[tj]: Thanks.. I don’t have a jail but I’m assuming I can use the vm as a jail.. also after I create that bridge and bar so I still add vm switch create public then cm switch add wlan0 ?

Vm*

ivy: I've bridged wifi interfaces before using the same process as I use for wired interfaces. Haven't had a problem.

Granted it was *only* bridging. Nothing special after that. No NAT or anything.

perhaps the wireless driver is rewriting the MAC address to be its own address, which would work if you're only bridging local interfaces (for jails etc) but not if you want to bridge an actual ethernet into a wireless network... if the latter setup works by default i would be surprised

oxbar: an epair and a tap here are equivalent and it will work

oxbar: but also, read the entire thing it covers a vm in bhyve

I trying to install xfce with pkg install xfce, but only got 'No packages available to install matching 'xfce' habe been found..

running Fbsd 14.2

nwe: pkg search xfce

ah looks like the metadata package is named xfce4-wm :)

nwe: i think that might just be the WM, not the xfce4 meta package

Grabunhold_: I'm not sure how, but the problem magically went away

I went to test with usb tethering instead of ethernet, that worked

then I went back to the ethernet and it still worked

I guess I'm just gonna cross fingers that it'll never recur

Grabunhold_: aha, but I cant find any other packages that looks like xfce4 meta package.

is there a way to tell zfs list to output all values as MB or GB so they are consistent? i know -p outputs bytes, but that's too big ;]

all values in gb, not 'whichever human readable' ?

yep

like df -g

i do not recall that being an option

man zfs-list doesn't indicate anything

what would be kinda useful is a little program you can pipe stuff like zfs list into an convert the units to the units of your choice

surely someone must've written something like that at some point, it's very unix philosophy

closest helpful thing might be zfs list -S used (or whatever) to sort by size, so at least you aren't going back and forth randomly between K/M/G/T

scoobybejesus_tl: that does help

"human readable" breaks so much stuff

even for humans

i still scream on redhat boxes where they don't have df -g

how do you make 'kyua test' just print why the test failed instead of writing it to a database?

try kyua debug instead, maybe

mtll: that sort of exists for zpool list

(my script adds hours/temperature to the drives)

kevans: ah that's a bit better, thanks

ivy: I think there's some ongoing work to make kyua debug more useful for debugging

creature comforts like giving you a chance to debug with artifacts on the fs before it nukes everything

any ideas on this build failure? pastebin.com/pgvrDnaB it seems that py311-wheel044 and py311-wheel are both pulled into the dependency tree

it seems py-psycopg2 depends on py311-setuptools, which in turn pulls in py311-wheel044. but py-psycopg2 itself pulls in py311-wheel

leading to a clash later on

note to self ... the packets do not go through the firewall if the wifi interface is down

anyone tried to build github.com/koreader/koreader in freebsd?

dch: how long were you looking for that problem before the lightbulb moment happened :P

polarian: both before and after dinner

yikes, not as bad as this bugs.freebsd.org/bugzilla/show_bug.cgi?id=285833

long enough to start pflog on the server

couldn't figure out the issue, had a sneaky suspicion what it could be, I believe it was i vy which reassured my suspicions that the docs were wrong :P

it caused me to procratinate for 1 month :P

installing py311-setuptools via pkg before attempting to install py-psycopg2 from ports fixes the issue with py311-wheel044

fold: I don't see why it would not, lua works just fine on freebsd afaik, I assume it just hasn't been ported yet, you could give it a shot yourself? has a nasty licence though

polarian: yeah I just want to set a laptop to read and take notes

and I was thinking of using freebsd

maybe I should try then and report

it's mentioned here what to take in account github.com/koreader/koreader/blob/master/doc/Porting.md

I've never ported anything to freebsd so maybe you understand better if that could happen? polarian

19:22 [ Demosthenex] i still scream on redhat boxes where they don't have df -g ...wait until you realize redhat has default aliases for rm/mv/cp to include "-i"

and while we're at it, try to use --group-directories-first with ls... ;-)

fold: actually there appears to be Linuxisms within the C code

it also seems to be GNU makefile which sucks too

BSD make is different

github.com/koreader/koreader/blob/master/Makefile#L17-L25 also checks for Linux

and apropos linuxism, bsd ls doesn't have --group-directories-first but --color ;-)

I didnt see the C code because they have split off the code into different repositories

I am not a porter so I cant advice on it sorry...

hmm, "pkg install caddy" -> "No packages available to install matching 'caddy'"

it's still in ports, and I installed just yesterday. where did it go? will it come back? was there any decision to remove it? how can I find out what's going on?

fold: I do suggest trying xournal++ though

handwritten and text note support

its within the freebsd port tree if you want to try it

its what I used before I finished school

oh wait

I have misunderstood what koreader is lol

xournal++ is quite different in that regard...

yeah koreader is just a book reader

the good thing is that I use that on android,kindle,linux

I wanted to use it on freebsd

but seems like a lot of things depends strictly on linux

it happens, some thing are just too linuxy

maybe using the linux compat?

that's worth a shot

anyone? any way to know why the "caddy" port has no prebuild package any more as of right now?

just set up FreeBSD 14.2 with DWM on a Thinkpad T490s, working nicely.

Grabunhold_: py-setuptools only build depends on py-wheel044, so first building that and later py-psycopg2 should work (or building in clean environments)

armin: nice! I'm waiting for the arrival of a carbon x13 gen 1

or something like that

I want to try niri in that one

and sometimes stuff fails to build (like recently go) and all that depends on it gets skipped and aren't part of the pkg repo as a result of that

nimaje: "pkg install py-setuptools" beforehand did the trick. all of this stuff is part of my ansible playbook, and after fixing the pgsql installation now caddy breaks =/

yep, go failed to build on the official builders, so everything written in go (like caddy) got skipped

nimaje: do you have a link to the build logs? i'd like to have a place i can go check when i run into surprises like these

portsfallout.com and pkg-status.freebsd.org

fold: Well, quite frankly put, I lost faith in Thinkpad's :-( It's sad to have to admit that, but I bought a T490s 6 weeks ago and I'd say it wasn't worth the money even though I paid only 300 bucks for it. These things are plastic bombers in the todays world and it's faaaaaar from the build-quality I was used to in Thinkpad's to say the least...

fold: It runs FreeBSD just fine so I'm more than happy, but still, the decay in build-quality over the last 10 years is not to be neglected...

fold: so yea I would say get one if you can get a used one for cheap, but other than that, ewwwwwwww, difficult to even justify the purchase compared to the Macbook's I got over the last 10 years, yeah.

fold: Still, it works nicely with DWM, it runs pretty solid even, yes slower than Linux but as long as it

...'s all scripts I know how to get that stuff going...

(sorry, ISO-DE typer here, the #/' key is next to enter here)

nwe: I think it might be called "MacBookPro11,1 (A1502, EMC 2678)"

I got a 13 (!!!) years old Macbook Air here that runs Linux just fine, I'd be surprised if the trackpoint support would even be halfway usable under FreeBSD (hello 3-finger-drag...), but I might try it

nimaje: thanks!

armin: i bought a used x13s a few weeks ago and i'm quite happy with the hardware so far (now the firmware is another story...)

Grabunhold_: well congratulations, I love my T490s, too, but compared to literally all the Macbook's, even the 13 years old (which I just cleaned entirely, looks quite nice) I'd say the build-quality just can't hold up...

armin: well, be that as it may, the non-standard keyboard layout and the firmware on the macbooks is a dealbreaker for me

Grabunhold_: that being said, I'm pretty much an Apple fan, I must admit that... yes, both hardware and software...

Grabunhold_: not much of an issue, you get the thing you're used to back with Linux (by default) and you can even switch to that on macOS

Grabunhold_: but I get your point, took me a couple of months, too

Grabunhold_: Muscle-memory dies HARD, I got to that problem the hard way, being used to ^A as a prefix from screen, coming to tmux (hello, ^B)

Grabunhold_: there's this thing deeply rooted in your veins that you need to be cautious about, I totally agree, if you have something in muscle-memory, definitely configure your environment exactly so you have it again.

armin: i decided at some point that i just don't want to deal with the muscle-memory problem, and fighting the firmware to let you install your own OS seems to be an uphill battle. so, thinkpads it is.

Grabunhold_: sure, there's thousands of valid approaches here, if that works for you, all fine.

Grabunhold_: but meh, I type ISO-DE since 1985 now, there's NO way I'll go ANSI

armin: fwiw, it's pretty easy to remap tmux to use ^a (i didn't want to fight muscle memory, so it's "set -g prefix C-a" for me)

afaik they used ^B for tmux, because they developed it in screen and wanted to test it easily, so the default prefix couldn't conflict

cyric: yup got that one, too, thank you. :)