hi everyone !

could someone help me understand why when I try to pkg install certain things it tries to uninstall all of my hyprland pkgs?

hm

now you have me curious if mine will pull the same thing

let's see :)

today I tried to install vlc and libreoffice and it tried this

pkg: sqlite error while executing grmbl in file update.c:154: NOT NULL constraint failed: packages.path

oh... well.. how about that

vlc and/or its dependencies conflicts with hyprland and/or its dependencies, should be "simple" as that

well last time i was able to resolve this by specifying those hyprland pkgs in pkg install ...

and it didn't try to remove them

i've had chains of pkg uninstalls from installing e.g. php packages

maybe it's the hyprland-qutils

oh. i guess it's only quarterly

so changing to latest you don't see it

yeah i changed to the latest

there's a bit of irony to that :)

yeah i had issues yesterday with kitty

well.. i mean.. i'm just saying that the quartly repo is kind of designated as the stable repo.. but then again that applies more to packages than pkg itself

yeah i don't know what the qualifications are for "stable"

quarterly seems just arbitrary tbh

not really. it seems like a standard practice to have a 'tested and true' repo

quarterly is just a snapshot of latest at a particular time, then it only receives security fixes and notable bug fixes

that's what i mean. it doesn't imply stability by being a time cutoff

i guess that would depend on how quickly pkgs are pushed to latest

and bugs are found

no, it's more stable because it doesn't receive any major updates, so things are less likely to break than in latest

that doesn't necessarily mean nothing breaks though

it should be entirely possible to have a local AI box manage all these builds and testings and designate things stable based on outcomes of the patching and testing

as long as it's not gemini it should work i mean

lol

ok, i solved my pkg conflict the same way as before. i'm not sure why pkg thinks i don't want these pkgs.

i've been away from ricing computers for so long that i forgot how useful tmux is when you're doing this sort of stuff

I live in tmux these days. Saves me a lot of hassle bouncing between computers and shelling everywhere

works well when you have an old tiny screen thinkpad too

pkg-lock(8) might help too: pkg lock is used to lock packages against reinstallation, modification or deletion.

oh ok, thank you

I have used it in the past to prevent upgrades of weechat that I compiled from ports with custom options

i'm not sure if the trouble of learning hyprland is worth it

my memory just isn't what it used to be. I think i've googled how to open a terminal like 3 times now ffs

intalled qubes os yesterday on a lenova desktop box

i need to bring in all my old shell and vim config files. i still haven't gotten to that. i just reinstalled freebsd after having gone nearly a decade of not using it.

qubes is wonderful. i had that on an old ibm rack computer a long time ago.

its my first time

proxmox though...

if you have the cores and memory it just makes all kinds of sense

I have to go trought documentation which is very well made

anyway just sharing

that always helps

so i have freebsd on my laptop and qubes on my desktop to play with

ivy: posted a patch to 292232; I will test it ~tomorrow, probably, unless you have time before that

kevans: thanks, i'll see if i can test it later

*nod* sounds good

probably needs more to makek the failure mode better than 'nothing happens', but i'll take the 'make it functional for now'

as long as you can install the system and boot it that is enough for now :-) i'm still not sure if we want to add kernel= to loader.conf in bsdinstall

well, i don't know if i had commented anything contradictory on this, but in tonight's opinion we should if it's not 'kernel'

there are very good reasons to turn off kernels_autodetect, it'd be better for loader.conf to be at least somewhat functional without it

i don't think anyone was really opposed to it, it's just a bit of a departure from how we used to do things, but it's probably unavoidable especially if we want to start shipping multiple kernels on the media (debug/nodebug)

once upon a time i might've argued that it wasn't necessary because kernels_autodetect is supposed to do the right thing

trying to run linux-sublime-text, but getting error "Unable to load libcurl.so"

How to fix?

Chip1972: you need to instal that library (for linux, not for freebsd) in /compat/linux/lib (or similar directory)

kevans: tested the patch, all working

beastie: how to instal that library?

hmmm.... don't know.... i'm not expert on linux distribution that comes with freebsd.

a lot of common linux libraries are provided as ports, i don't know if curl is one of them

try to search a package named linux-curl or simiar.

if "linux-sublime-text" is a port, it should probably already depend on that though, so that would be a bug

there are two such packages: linux-c7-curl (centos) and linux-rl-curl (rocky linux)

ivy: you are right, but the package owner should make the package to request the dependency, that simply doesn't happen automatically.

Chip1972: for the next time, i've just done pkg search 'linux.*curl' and got both packages

I found, and they already installed

probably you need to do an ldconfig (the linux one) to rebuild the shared libraries database. Try to find if the library is actually in some place under /compat/linux

yay! pkg mess is fixed.

found libcurl.so.4

where was it found?

probably you need to run the linux ldconfig tool to add that directory in the scan. Probably it's not finding it because it doesn't find the library in the database.

found libcurl.so.4 in /lib and /lib64

IMHO linux and freebsd have different versions of ldconfig, to build the libraries database, as the freebsd requires the directories to search for libraries in the command line, while linux follows a different approach (there's a configuration file that states which directories to search) you should search for a ld.so.conf or ld-so.conf file in the linux

i don't actually know if that's included in the freebsd start-up (probably the package installs in /usr/local/etc/rc.d/something)

I just following the tutorial from freebsd site

my /compat/linux/etc/ld.so.conf just has an include of ld.so.conf.d/* but that directory is empty.

O will try run Alpine in chroot

Chip1972: I'm telling you this because linux is linux and this feature, despite being very similar, is not exactly equal. I started using freebsd some ten years ago, but I started using linux in 1994, so i know better how this happens on linux.

that will not help.

I dont know what is ldconfig

running chroot will not make ld-so.so (or whatever the name is) to find and load libcurl.so

ldconfig is the tool that builds the database of shared objects installed in the system to be able to find quickly a shared library.

but man ldconfig will explain that better than me.

the system libraries are not searched by a PATH like strategy, but are indexed (due to the number of installed libraries and the repetitive of the process) in a file that needs to be built at startup.

well... I works with electronics, studied physics, and now studing biochemistry...

computer is a tool who demands too much effort to do things properly

since zfs arc will kinda expand to use whatever free mem there is, then shrink as other things actually need it, how do you guys find out the actual free mem of a system? free meaning that which can be freed up and made available to things that need it

the mem line of top has a "inact" field, how would i get that from the command line?

beastie: how to run ldconfig on /compat/linux?

kerneldove_: vm.stats.vm.v_inactive_count, it's in units of pages so multiply by page size (4KB on most systems)

is there a sysctl to get the page size so i can make it dynamic in my code?

ah yea

vm.stats.vm.v_page_size, nice!

tyvm!

ivy am i right in that inactive mem is basically free in that other things can use it without causing the system to crash or swap or other progs to terminate from OOM?

ah yes that's right, confirmed by allan jude himself unix.stackexchange.com/a/137254

Have any of you used OPNsense on 15?

where does systat -ifstat get its "load average" figure from?

kerneldove_, sysctl vm.loadavg

why is that displayed in the output for network stuff? (-ifstat) isn't vm for mem?

kerneldove_, The answer to that is in the man page for it. "The upper window depicts the current system load average. The information displayed in the lower window may vary, depending on user commands."

ahh ty! so system load average refers to the vm load average. for some reason i always though system load referred to cpu usage but it's actually vm subsystem

so it seems my server is fine until it hits about 29 load average, then it starts dropping network throughput pretty substantially

is that normal?

load average is the number of processes currently running + the number of processes waiting to run

load average is a quasi moving window average (to 1m, 5m and 15m) of the number of processes in the R state (this is, ready to run) in the system.

what do you mean by 'vm' ???

well if load avg isn't specific to vm, why is it under the vm mib in sysctl?

vm means virtual memory no?

it's not system load, the system load depends on many factors, in this case is the number of total threads in state ready to get the cpu.... if you have three cores, and you get a load average of 7 you can infer that the number of threads ready to run per core is about two

vm meaning virtual memory means nothing to load average.

beastie: i think kerneldove_ is asking why the sysctl mib is vm.loadavg when it's not really clearly part of the VM subsystem

yep

a process needs only one page of code segment, one of stack and one of data segment to be able to run in a virtual paged memory system.

It's probably just a historical artifact. It needed to go somewhere and really could go anywhere.

probably it's not part of the virtual memory space... not.

kerneldove_: fwiw, i think the answer is that vm is broadly "the part of the kernel that deals with processes", so a lot of stuff gets lumped there

you talk in riddles beastie, it's really not useful fwiw

huh weird, ty ivy and rwp

loadaverage is a term coined from the program uptime (which should neither do the load average calculation)

kerneldove_: I don't understand what you mean with riddles, let me translate to my language.

no, I try to be precise in my answers. I've mathematical background, sorry if it disturbs you.

i try to help, not to annoy.

'lol'

is 16G a good swap size for a server?

i try to keep swap unused but i think it's good to have for safety?

kerneldove_, The amount of swap needed depends upon what you are doing. What's your server serving? Probably it does not need that much. But with disk sizes in the multiple terabytes using 15 gig of it for swap is not otherwise noticed so okay.

tons of web serving basically

ya like i said i try to balance workloads so swap is never even used

you need enough swap to store a kernel crash dump, which depends on how much memory your kernel typically uses, but a couple of GB is usually fine

For that then you probably don't /need/ 16GB but it won't hurt anything.

128gb physical ram

Sounds good. Go with it. It's fine.

why do i need to store a kernel crash dump?

If and only if you run into kernel crash problems then the kernel crash dump can be used to debug it. If you have never had a kernel crash and don't expect to in the future then this is just an unused but available feature.

i notice default debian installs only have 1gb of swap and those are used as servers all the time

and also debian is way more eager to swap than freebsd is

It is good to have at least a little swap. It allows the kernel to free up memory that is never going to be used again and use that memory for other things. Memory is valuable. But then you have 128G of it so...

with Linux you can set "swapiness" and it makes it more or less prone to swap things

Know I'm going to get judged for saying this. But, I do some swap on my systems to keep them running if they run out of regular ram instead of just crashing

And both of my systems have 128G also

how much swap you set?

Also don't confuse having swapped out some pages that are never used with swap thrashing due to not having enough memory to support the resident set size of a process. Two different cases.

kerneldove_: I have 24GB of swap on this machine, and likely similar on the other -- It is literally what Alma Linux just did for me when I told it "automatic partitioning" lol

wow

how much of that do you see used at most?

Looks like 8G right now usercontent.irccloud-cdn.com/file/hI7PGkDN/image.png

what's your load avg?

Depends if my FreeBSD vms are being used to build things :)

I have most of my cores, and half of my ram dedicated to them

nice

bet you're glad you got 128gb before ram prices went up 5x lol

I've ran without swap before too. But that was on my machine that had 256G in it :P

A friend posted this image recently and I think it is worthy to share. 128 cores all busy doing builds. htop.dev/images/1

28.png

ya i have some systems with no swap too

wow

90 load avg damn

kerneldove_: What are your system specs, and what do you normally use it for?

mostly ryzen 9 3900x 128gb ram 2tb nvme, servers

kerneldove_: the reason you want to store a kernel crash dump is because a kernel panic can be harder to debug if you don't have the dump

what do you serve up with the "servers" ?

web pages and stuff

ivy ok ty

What web server(s) do you use?

I've fallen in love with the simplicity of Caddy

ya i use caddy too

used to use nginx

i guess i'll use 16gb swap for the new server configs and just start scaling down workload if i detect 25% of it or more is used

Yeah, I did nginx prior also. But with Caddy I do a couple static web pages, and a files listing to load my User Manuals -- and as a reverse proxy for Plex, Jellyfin, and a pastebin

i liked how it auto set up lets encrypt, but i really hate caddy's logging config

for what little I do, Caddy does it all easier than anything else

pkg repos were broken for several hours but is now working again

bad

i just set up a few new freebsd servers at a hosting company. i set them all up the same exact way with scripted bsdinstall. i can now ssh into the 3 servers from my freebsd computer, but when i try to ssh in from my debian computer i can only connect to 2 of the 3

the 1 i can't connect to says connection refused. i tried disabling pf on the server but error was the same connection refused

any clues?

first time i've run into this kinda problem

ok nvm heh

Do tell, what caused that to happen?

kerneldove_: try to ssh to one of the servers that accepts and then jumphost to the one that refuses connection. /var/log/auth.log will probably give you an explanation

i put the wrong fucking ip in /etc/hosts heh. somehow i put a 6 for a 9. some kinda vertical dyslexia?

wait, I think we _can_ boot off a temporary root fs. Huh

in other news despite labelling the pi as "stop playing with this" I have nevertheless loaded uboot.bin onto it using rpiboot, then ymodem'd loader.efi over the serial port and got _that_ running

so now I have a raspberry pi booted that far without an SD card in it

I think a bigger uboot build would allow me to yeet the efi loader over the USB connection at far greater speeds

somehow I never learned of rpiboot so far...

pretty neat, actually. It's quite handy for tinkering with bootup stuff without constantly pulling the SD card

now the real question is how to get some kinda image into memory that u-boot can read, such that it can be tickled into booting freebsd entirely into an md

that'd be pretty cool

sometimes I forget you can do uboot on a Pi at all...

of course uboot does 1,372 things I don't know about. I played around with it on a RISC-V board where you could build first-stage bootloader from uboot, and that was at the level even of things like DRAM training, which handed off to OpenSBI, which then handed off back to another uboot stage

A freshly built go binary that's semi-min in design file(1) output is "for FreeBSD 12.3" is this FreeBSD's way of saying this should run on 12.3+?

I see other that are 14.3 (the host) but are linked to a bit more (vs the above).

Is that go that making it backwards compatible or is that a FreeBSD thing? Looking at all other bin/sbins it seems everything is for 14.3

I looked at two go binaries here, and one says 12.3 while the other says 15.0

The first one I built from ports, the other I built with poudriere

if file(1) says 12.3, it should run on both; however, you may need the compat12x-xxx package

I can say for mine it's not needed.

Best I can tell the ELF header desctipion is populated with 123000?

That at least answers why file says that

Hello! Has anyone managed to configure the aerc email client to be used in combination with protonmail? (Using hydroxide as protonmail replacement, since the official one isn't suppported on FreeBSD yet/hasn't a port yet)

lockna I use it with thunderbird

And hydroxide is probably better than the bloat that is the official client, anyway

looks like certificate of events.eurobsdcon.org is expired.

Running the pkgbasify.lua script now <fingers-crossed>

aaand it worked.

I hope you updated to 15.0 before running it

lts: Yep, running 15 for a while now

wavefunction: how long does it take ?

mzar: About 15 mins with all the installs,

Err, 10 mins *

thanks

I have a relatively nice fiber internet connection tho.

Downloaded something like ~313 packages

has anybody compiled ungoogled-chromium locally?

very nice little tutorial on how to create a minimal bootable FreeBSD: youtube.com/watch?v=Uc8Z-XX1TS4

can you increase the performance of a freebsd server much by customizing the kernel and compiling on the hardware it'll run on? or is it negligible?

performance? not really. maybe free up a small amount of memory

my custom kernel was i want to say half the size compared to the .. 28MB generic kernel

kerneldove_: you can increase performance significantly by adjusting some sysctl variables, I tink it's covered in Handbook Chapter 12: freebsdhandbook.com/config

ooooh, neat video! I had no idea it could get _that_ minimal

having discovered today that we do have hte benefit of something like an init ramdisk I do want to give that a go at some point – perhaps building a system that's juuuuust enough for logging in and loading a decryption key for zfs or geli, perhaps