i think i found a bug in bsdinstall or its man page but not sure

DISTRIBUTIONS says the default is unset, but it's not really because if you leave it out, kernel and base still get installed(???)

and where does base.txz get pulled from? asking because if i gotta add my own conf files to the bsdinstall script DISTRIBUTIONS env var, i gotta know where to put my.txz

usr/freebsd-dist/ has it so can i put my.txz in there too?

i don't wanna step on the OS obv

polyex: what version?

13.2

and image/arch etc?

disc1 iso amd64

hmm well im a newbie but since noone is saying stuff, it always pulls kernel and base for one

do you have any logs?

!paste

oh its in the banner

im making a scripted bsdinstall you know what that is?

after install, why don't i got scripted bsdinstall's log at /tmp/bsdinstall_log like man bsdinstall says it should?

Is that after rebooting into the installed system?

ya

I haven't used it myself, but from the manual page, I suspect the log was (and may still be) in whatever /tmp was while bsdinstall was running, not the /tmp of the installed system. If bsdinstall's /tmp wasn't persistent, it may be gone.

how do i read the log then?

If it was on a filesystem that persists across reboots (and wasn't erased during your scripted install), it's probably still there.

well can i somehow get cli access to the newly installed system before first reboot?

You can do whatever your scripted install lets you do. You wrote the script and can presumably modify it. Or is that a different convo?

well after scripted bsdinstall runs, it automatically reboots

can that be disabled?

i added a line to bsdinstall's script section to cp /tmp/bsdinstall_log to /root/ and before the first reboot i saw console output that the file wasn't found!

From bsdinstall's manual page, "SETUP SCRIPT" section: "This script will be run at the end of the installation process *inside a chroot(8) environment* in the newly installed system" (emphasis mine). So the log is probably outside the chroot.

ok so if there's no way to get to the log what use is it? i gotta be missing something

Did you call your script "/etc/installerconfig"? That's the only place in the manual page where automatic rebooting is mentioned.

ya

Did you try calling it something else and running it using "bsdinstall script /path/to/install/script"?

no, why would i do that?

what do you think is going on?

Heya. So, I have all these extra buttons on my mouse. Can I connect UI configuration changes to them? Like say, triple the mouse speed when I hold down one of the two thumb buttons?

Or, how might I affect mouse speed from the command line? I don't mind running a background script that parses output from `xev` if I can figure out what to do when I see the events.

polyex, I think you missed part of the paragraph (just above the PREAMBLE section) that reads "On FreeBSD release media, such a script placed at /etc/installerconfig will be run at boot time and the system will be rebooted automatically after the installation has completed."

no i know that

so how do i read bsdinstall_log

So by calling something else, you may be able to avoid the autoreboot at the end.

i tar up a dir called config, then in distributions i add config.txz. on the finished system config.txz was extracted to be /config/... how can i not have it be at the root?

got it

needed tar -C

after a scripted bsdinstall, should i have it run freebsd-update fetch/install to upgrade to any latest patch level on -RELEASE?

what's the best way to shorten the boot delay from 10 seconds to only a few?

autoboot_delay in loader.conf i guess

how do we inject a line of text into a file only if it isn't already in there?

grep -Fqx 'line' 'file.txt' || echo "line" >> file.txt

can i make it any better?

Look at sysrc polyester

Erm polyex damn you autocollect

autocollect?

and look at firstboot cgit.freebsd.org/src/tree/release/tools/oci.conf#n54 to see how that is used

Title: oci.conf « tools « release - src - FreeBSD source tree

autocorrect

I’m on my phone atm but AMA later today

TLDR use firstboot daemon and touch the /firstboot file for this to run

i'm totally lost

Use that to install/update patches, and packages, to keep your image size down and not need always to be rebuilding it

look at sysrc for what reason?

Sysrc will replace your grep |echo thing to set a flag in /boot/loader.conf for example for auto boot delay

service firstboot enable ; touch /firstboot

i can use sysrc for more than just /etc/rc.conf?

yup it just wants a path to the file you want to modify see the man page for details

wow ok!

so i use that for text injection but what's all that service firstboot stuff?

i think you're saying do my configs on first boot. but why wait? why not get the whole config in place on installation so there's only 1 reboot

up to you and how often you want to rebake your images

why does your system make it less often?

I bake the image once and on first start it installs all the patches and up to date packages

I guess the main difference is I deploy the image directly without using the installer

ok so i bake the image once and after install, before rebooting, it installs all the patches and up to date packages no?

So there’s no other place to install stuff in my model

oh hmmm

Either way is fine

how do you deploy the image directly without using the installer?

dch

polyex: ipxe boot to a linux shim that does dd

polyex: i would prefer to use mfsBSD for this but i didn't get the uefi boot working from ipxe for that

ah ya very nice setup

pxe is cool

looking at man page but don't see it. how do i use sysrc to insert a line in a file if it doesn't already exist?

yeah, mfbsd missing uefi is a huge downside

stupid question, is rm -rf /some/dir atomic ?

now i wanna try it

ok wow

most commands not found, but exit worked. wouldn't boot

rm -rf /*

mage: no it needs to traverse leaf files and remove them first. if you rename it first maybe that helps your use case?

yeah, so if we're on the same fs I could mv old_dir new_dir ; rm -rf new_dir

Title: Snippet | IRCCloud

^ polyex

anyone knows where transmission-remote command went with the new ports?

the cli version

sopparus: what'd that do?

its just connect to transmission-daemon and do stuff, like stop, start list etc

oh, is that a bittorrent … thing

i think it was included in old transmission package

yes

pkg provides bin/transmission-remote shows one in transmission-utils, and one in transmission-default

yeah utils wants the daemon uninstalled

ill mail the maintainer, i think its wrong

thanks

You could also just create a bug, which… does the same, but is tracked.

ah it was allready on bugs bugs.freebsd.org/bugzilla/show_bug.cgi?id=274065

Title: 274065 – net-p2p/transmission: -daemon and -utils still conflicting after PR#273841

273841 – net-p2p/transmission: -utils and -daemon conflict after update to 4.0 bugs.freebsd.org/bugzilla/show_bug.cgi?id=273841

Hello, its me again :-) We have now set up a new 12.4 build host, checked out the 13.2 source under /export/fb13/usr/src and wanted the start the buildworld there.. That works well, but it always takes /usr/obj as OBJDIR (the default). What we want is /export/fb13/usr/obj. So i tried MAKEOBJDIRPREFIX=/export/fb13/ as env var, which worked well, but it still appends CURDIR to the path so it results in /export/fb13/usr/obj/export/.... I've

also tried MAKEOBJDIR which results in a build fail.. Am i just to stupid or isn't it possible to strip away the curdir and set an "absolute" path?

what do you mean by "absolute" path?

yuripv My goal is to have "/export/fb13/usr/obj/usr/src/amd64.amd64" instead of "/export/fb13/usr/obj/export/fb13/usr/src/amd64.amd64". So strip away the 2nd "export/fb13". I tried with build like this: MAKEOBJDIRPREFIX=/export/fb13/usr/obj/ make SRCCONF=/export/fb13/etc/src.conf buildworld -j24

setting MAKEOBJDIR to /export/fb13/usr/obj/${.CURDIR:S./export/fb13/..} might do it

hm, no, probably won't work

setting MAKEOBJDIR broke the build completely.. and isn't really supported according to: man.freebsd.org/cgi/man.cgi?build(7)

Title: build(7)

well, there's always the sledgehammer solution: nullmount /export/fb13/usr/src on /usr/src

Hehe.. yeah that would be an option.. but not really a nice one :-P

so I think this might work: don't set MAKEOBJDIRPREFIX, but set OBJROOT to /export/fb13/usr/obj/usr/src/

You just read that from the makefile, as i guess this isn't really a documented way?

yeah, see src.sys.obj.mk

personally I'd use a nullmount :-)

That does seem to work

OBJROOT i mean

Hi all, this morning when I tried to boot my systems (FreeBSD-13.2, GELI+ZFS on SSD) I got error 22: Invalid fstype, so I can't boot. I created a live USB and booted into USB. gpart shows all the partitions (both the USB and SSD) but what's next? "zpool import" gives "no pools available" so I think I'm missing at least one step. There were no

upgrades on the system in the past several weeks and yesterday the system started just fine. Any ideas? TIA.

bolik: you have to decrypt your devices prior import

megaTherion I do get the promt for the geli password before boot, that goes well. Is there additional step there?

I don't think so, geli devices would appear in /dev as .eli postfix (I think)

but that is the same what gpart show's I guess, is zpool status showing anything?

zpool status shows "no pools available", which makes sense because if there's no geli device, then it cannot see any pools... at least in my simplistic reasoning

but then why is there this "error 22" in the first place? Is this something to do with the drive, which is possibly corrupt?

bolik: at what point in boot did you get the error?

bolik: and have you tried doing a geli attach from within the live USB?

RhodiumToad halfway through the boot process (or you mean more specifically?)

will try it, just need to find the correct command

yes I mean more specifically - where is "halfway through"? what messages had previously appeared?

geli attach did it. will try to import zpool

import zpool -R /mnt zroot gives: cannot import 'zroot': no such pool or dataset. Destroy and re-create the pool from a backup source

zpool import without any options says what?

RhodiumToad shows the name of the pool: zroot and its numeric id, state: ONLINE, and that tha pool can be imported using its name or numeric identifier.

zroot: oNLINE, nvd0p4.eli: ONLINE

but importing it fails?

yes

show the exact command and error

RhodiumToad i guess the nullmount would have been the better way :-P when mounting the built source via nfs on an other server and trying to install there the paths, vars and so on must be adjusted too.. so i guess null-mounting would be better to have a default setup...

the exact command is: "zpool import zroot". The output of it is: cannot import 'zroot': no such pool or dataset. Destroy and re-create the pool from a backup source.

and the same if you import it by id?

yes, the same error

hm. I don't know then.

"backup source" sounds fine

but never had such an issue either

bolik: what's file -s showing on the eli device?

but wait, I think that really doesn't work with ZFS anyways - there was a better way

wasn't there a zfs debug tool or something

there is also a recovery option for zpool-import

something like zdb?

ya but Im not sure how that works with non-importable pools

that's a good question, don't know that.

damn connection...

running sort -i -n -k 2 on the output or irssi's /list gives "sort: mbrtowc error: Illegal byte sequence". is there a way around this?

*output of

LC_ALL=C sort ... perhaps?

yes that worked, thanks

the error means that you have LANG or LC_* set to some encoding and the data you fed into sort isn't in that encoding

it looks like I have LANG set to C.UTF-8

LC_ALL in the environment overrides all the other LC_* and LANG variables

ok, so the error is the result of the input not being valud utf8

*valid

yes, it appears so

rsjw: i'm curious. what is it?

oh, irssi's output

does irssi still default to ISO 8859-1?

irc is notoriously loose about encoding

I think it's just what weird nonsense people put in the topics

I'm trying to change mouse speed from the command line. `xinput list` tells me I have "System mouse" at 6, and `xinput list-props 6` tells me "libinput Accel Speed (289): 0.00". But I don't think that's what I want. What changes the mouse speed?

Is it "Scrolling Pixel Distance"? Am I looking in the wrong place?

any PC Engines APU6 users around? I tried FreeBSD on mine with pf and the network throughput went down from 700 to 300 Mbps when using a really simple pf rule set

vortexx: what's the really simple pf rule?

and what sort of traffic

APU6 is not a particularly fast box but should come with good NICs, but that seems a bit on the low side assuming this isnt pathological traffic

I'd look to see if its hitting any cpu or nic related limits

meena: I used this /usr/share/examples/pf/faq-example1 this with my interfaces set up.

dch: using iperf3 for testing, and speedtest-cli

dch: the nics can drive close to a gigabit of traffic with a basic install doing nothing and testing with iperf3. They're I211s

Did anyone who subscribes to freebsd-stable get this one? lists.freebsd.org/archives/freebsd-stable/2023-September/001485.html

Title: Fusefs unionfs trouble with hardlinks?

I got dozens of DMARC errors in response, but it did show up in the archive

hey

how do i use sysrc to insert a line in a file if it doesn't already exist?

...just use sysrc as usual? It'll add if not there, modify if there..?

a file that isn't /etc/rc.conf

hm maybe it's that easy

doesn't seem like it

sysrc -f path/to/file line, says unknown var line

i wanna be able to ensure an arbitrary line is in a file i pass a path to

touch the file first?

it exists

-i ?

tried it

didn't complain but didn't add line to file

what do you want to happen if the line is, or is not, already present?

if present do nothing. if not present add line

add it with what content?

just to ensure a line, is in a file

the line

string

text. dunno what you mean

sysrc doesn't do "lines". it does var=val assignments only

ahhh

now i get it. ty!

sysrc foo=bar replaces any existing foo=whatever with foo="bar" and so on

sysrc foo returns whatever foo= assignment exists, or a default one, or whatever

so cool to have a command interface to modifying a text format's content

structured data ftw

well it's kind of a hack (but then rc.conf is kind of a hack too)

hack why?

rc.conf is just sourced as a shell script

so it's possible to do stuff in it that sysrc can't cope with but which rc.subr will happily process

why's that a hack tho?

not arguing jc

ah

maybe sysrc needs an upgrade?

it would be insane to try and make sysrc cope with arbitrary shell

ah ya

another q, after install, why don't i got scripted bsdinstall's log at /tmp/bsdinstall_log like man bsdinstall says it should?

shrug

wtf

i had autoboot_delay="-1" in /boot/loader.conf and the 10 second boot menu delay was gone. but then i moved the assignment into /boot/loader.conf.d/file and it doesn't work. why not?

why would it work?

what fbsd version and what loader version?

i thought loader.conf.d files would be automatically read in

13.2

how do i check loader versino?

what filename did you use?

loader

looks like loader only looks for /boot/loader.conf.d/*.conf and not for arbitrary filenames

oh duh my bad

sorry

i have a few dozen bhyve vms. i make a tap# interface for each 1 and add it to cloned_interfaces in rc.conf. it's kinda tedious. any way to eliminate that?

create them dynamically?

how?

I'm having an issue with the installer not mounting the ISO properly on my qemu virtual machine. Anyone know what I should do with this error? imgur.com/a/skNCb8H

your vm isn't emulating the ahci interface adequately

RhodiumToad: where would i go to fix that?

the vm does most linux distros and openbsd fine, is this just an issue with freebsd or did i mess something up?

no idea

are there any other messages before that?

i'll try getting it again, the screen is pretty small by default

how do i dynamically create tap if RhodiumToad?

ifconfig

before each use of the vm that uses that tap?

like before the bhyve command

RhodiumToad: I added another image, same link

polyex: a command like ifconfig tap create will create a new tapN interface and print its name to stdout

zayd: well, getting command timeouts from an emulated device seems unreasonable

then after shutting down vm destroy tap to clean up?

yeah

hm ya that could work. cause i wonder, would the system slow down if i tried to set like 10,000 tap# interfaces in cloned_interfaces?

jc

switching the cdrom device to usb instead of sata worked. odd.