meena, I was hoping you would like it. When I first read that I knew I had to try it.

If the file system can be reduced in size to fit into a ramdisk then the procedure all worked very well.

While I was playing with it I actually ended up screwing up a ZFS file system. So I went through it again to flip it back to UFS. And then again to do ZFS right the next iteration.

rwp: I'm down to 3.4 G, so I'll need to temporarily balloon my RAM.

i don't think I'll be able to get it much smaller than that

I do believe that it'll be smaller on ZFS, thanks to compression ;)

I was just doing it for the exercise so I started with a pristine bare minimum install. Which would fit into ram no problem.

In the blog posting the author copied everything of interest off to another system temporarily to make things small enough.

aye

It is a risk. If the system crashes while everything is in RAM then poof! It's all gone in an instant.

So it's good if the only thing on the system is the system itself and no data of note otherwise.

rwp: diskless systems do that

wild, bugs.freebsd.org/bugzilla/show_bug.cgi?id=269211

Title: 269211 – misc/unicode-emoji - Update to 15.0

we were at 5.0

cpet, Diskless systems?

What is that ?

cpet, You said: ":32 <cpet> rwp: diskless systems do that". Do what?

that!

This sounds more and more like the Meatloaf song by Steinman. I'll do anything for love but I won't do *that*.

:)

you won't, but diskless systems do

And the response makes no sense so I’ll continue to say what’s that

Does the freebsd API offer any non-root alternatives to the mlock() syscall?

What should I be looking for in order to update a "tar"d compressed file (via "xz" or "zstd") created from $HOME, to be stored on exFAT filesystem?

Or, a "rsync" of some sort to update the compressed stream

you're looking for something more complicated than `cp myarchive.tar.xz /mnt/` ?

Yes,

s/,//

even rsync will see that the new file is not the same length and recopy the whole file

i'm not sure what you're trying to solve by not just copying the whole file

who has v6 enabled working like a champ?

zidrav could make you a patch to copy over and apply instead

I think tarsnap works similarly, but I'm not sure

Which is the default version of Python in 13.1 RELEASE?

Is it version 3.9?

I run pre-build packages on my machine and see that most of the pre-built packages are of version 3.9. I'll stick to that version if you do not educate me otherwise :-)

thorre: I think so. you can check by looking at the dependency of freshports.org/lang/python3

Title: FreshPorts -- lang/python3: Meta-port for the Python interpreter 3.x

Thank you meena

Hmmm, version 3.3 is being referred. I will do some more reading up on this. Thank you for the pointer to freshports.

thorre: no. Runtime dependencies:

python3.9 : lang/python39

meena: Thank you for the interpretation, got a bit confusing there for a short while.

the version is lang/python3 has almost no meaning

s/is/of/

can I change a file's access / modification time if I have the desired timestamp in epoch format ?

or it must be full date, year, day, etc

last1: man.freebsd.org/cgi/touch(1)

Title: FreeBSD Manual Pages

Title: touch(1)

(almost)

saw it, was hoping for timestamp :)

ok, but I'm trying as they explain there, like so: touch -d file1 2022-11-18T15:50:24 and it says: touch: out of range or illegal time specification: YYYY-MM-DDThh:mm:SS[.frac][tz]

maybe the datetime has to follow -d

that would make sense, yes

last1: look at the SYNOPSIS

Hi

the only option that accepts a file is -r

is there a way to build a vm image on linux?

user062b: yes. our source tree builds on FreeBSD, Linux, and MacOS.

last1: you can also use date -j -f to convert from whatever format you have into whatever format touch expects. Or, you could just use touch -r to copy access times from another file

user062b: I think our build runs in GitHub actions (Linux) and on CirrusCI (FreeBSD)

meena:do you mean through "make vm-release" (after building kernel and world)?

yes

Title: freebsd-src/cross-bootstrap-tools.yml at main · freebsd/freebsd-src · GitHub

thanks, my main use case is to tryout changes in the kernel, but i haven't a way of changing the kernel on an existing image on a non freebsd host, are there other ways except building the image every time?

user062b: I use PkgBase.live to install all the kernels

I build every kernel that makes sense for a given architecture, and generally use GENERIC-MMCCAM.

meena: thanks a lot

user062b: you're welcome!

can't miss an opportunity to spread propaganda about PkgBase

i'm trying to write an rc script for a daemon. i've defaulted ${name}_user to my service user, but as soon as the rc script switches to the service user, daemon (i'm using it to start my service) is unable to create a pidfile because the service can't write to /var/run

i could of course create something like /var/run/foobar and chown it to my user

but i'm wondering if there is a standard solution for my problem that i'm missing?

hm, maybe using daemon's "-u user" facility

but that would require setting the user in .. ${name}_user2 or something so that rc doesn't eat my username variable?

Grabunhold_: i would stick to ${name}_user, and create a /var/run/foobar as pre-start. or you can create it in the port

you can sometimes also completely forego pidfiles: codeberg.org/FreeBSD/freebsd-ports/…t-im/gotosocial/files/gotosocial.in

Grabunhold_: maybe this will be helpful reviews.freebsd.org/source/ports/br…/main/sysutils/seatd/files/seatd.in

Title: seatd.in · ports

you can simply use command_args to pass any number of arguments to daemon

if I understand your problem correctly..

meena: looks like a really good template! where can I find docs for these %%PREFIX%% variables?

ngortheone: thanks for the template, but my problem was about the name of the ${name}_user variable (or the permissions problem with the pidfile, however you want to look at it...)

"System scripts should be installed from src/libexec/rc/rc.d through the Makefile found there."

but what if i don't have such a directory in my system

other directories named rc.d don't contain Makefile

There's a Makefile in /usr/src/libexec/rc/rc.d/ wouldn't be that file ?

Grabunhold_: those should be automatically replaced by the ports system

gasbsd: my /usr/src is empty, maybe i need to install a source pkg of some sort

Grabunhold_: variables not automatically replaced after in SUB_LIST and PLIST_SUB: codeberg.org/FreeBSD/freebsd-ports/…main/net-im/gotosocial/Makefile#L25

Title: freebsd-ports/Makefile at main - freebsd-ports - Codeberg.org

hubertm: what's you freebsd version ?

13.1-RELEASE-p3, userland p5

hubertm: try cloning the source to /usr/src, git clone --branch releng/13.1 git.FreeBSD.org/src.git /usr/src

Title: src - FreeBSD source tree

hubertm: what exactly are you trying to achieve?

meena: trying to add an rc.d script to be run during the boot

hubertm: which one?

meena: my own script which mounts an external hard drive

you don't need to install that from source. you can just place it in /etc/rc.d

Why not edit /etc/fstab for that instead?

or add a devd script?

meena: but something should be done in addition to placing it there?

hubertm: rc scripts usually need to be enabled

V_PauAmma_V: AFAIK fstab has no option to disable file system check if there's something dubious. it can take 20 minutes for booting in that case no matter if i want to use that drive or not

Title: fstab(5)

meena: thanks, i figured they might be Makefile variables by now (i've never written a makefile). at the moment i'm just trying to write the initscript, but i do plan to upstream it eventually, so that will come in handy :)

man.freebsd.org/rc.conf(5) also has

background_fsck

Title: rc.conf(5)

Isn't background the default?

yea

it pairs with journaling normally so it does not need to do the heavy foreground style check

in the meantime, daemon seems to have some kind of problem with the word "run"...

this works: daemon -S -T footest /usr/local/bin/ergo --version # prints the version to syslog

with this, "run" never seems to make it to ergo: daemon -S -T footest /usr/local/bin/ergo run --config /foo/bar.yaml

this has the same problem: daemon -S -T footest -- /usr/local/bin/ergo run --config bla

other arguments without dashes seem to work too, it's just "run" that seems to get eaten

I wonder if it's daemon, or rc.subr

Grabunhold_: can you post your rc script as it is now?

heh, i wrote one for ergo before but i didn't bother with chroot

what is ergo?

meena: yeah, gimme a sek, just came back to the computer again

ergo is an irc server, fittingly

rtprio: i'm not going to chroot either

i don't know why there's not a script in the port

Title: #!/bin/sh# PROVIDE: ergo# REQUIRE: DAEMON# KEYWORD: shutdown## Add - Pastebin.com

rtprio: well, somebody has to write it. so I might as well learn how to do it :)

when starting that script, ergp prints it's help text into the syslog

"Bleeding-edge IRCv3 support suitable for use as an IRCv3 reference implementation." nice

"High customizability via a rehashable (i.e., reloadable at runtime) YAML config." no. NO! NO!!

i'll pull mine out in a minute

meena: yaml engineering, it's all the rage these days

the worst Website, noyaml.com making the best argument

Title: 🚨🚨 That's a lot of YAML 🚨🚨

Grabunhold_: bsd.to/WJvc

Title: dpaste/WJvc (Plain Text)

rtprio: what is ${ergo_command}

daemon, in this case

it's at the bottom, line 42

that's mildly confusing.

yes, but i haven't looked at it since i wrot it

but it works

rtprio: thanks! but i'd still like to understand what's wrong with my config

looks like it should work?

if i substitute "run --config (path)" with any other command, ergo seems to get invoked correctly

if you get a weird egrep i might add a pidfile

Grabunhold_: you can run the rc script with set +x

oooooh it seems like it was all my fault, just a second...

yeah

it's --conf instead of --config

meeeh!

👍

i don't remember any other ircd config files being much better than yaml

rtprio: YAML is a reckless format. With m4 you know you're holding a straight razor. but YAML claims to be human friendly

Grabunhold_: do status / start and restart work?

meena: can't test, my yaml config is borked and doesn't parse :D

gotta fix that first

write it in json.

of course

json, while being awful itself, is a complete subset of YAML

iirc ergo uses features that are not in the json subset of yaml, the config is fairly complex

... i'm configuring the whole thing with ansible, so it's not like this is my first yaml ride haha. i'll eventually dump the config straight from an ansible dictionary i guess

by the way, while i'm not exactly new to FreeBSD, i'm significantly expanding the scope of things i do with it. just got started with jails (the irc server is a jail, too) and so far it's great

there was a learning curve of course, but i finally feel like the jails setup is gonna work for years to come. coming from docker and LXD containers on linux i really like how "boring" and "done" things seem to be

meena: hmm, puppet being somewhat like ansible in scope - what's put you off of configuring your jails from puppet? that's exactly what i do at the moment

Buildah? Is that something I should look into?

with ansible, of course

I've almost got LXC wrangled and boring. subuid assignments are still nettlesome.

meena has a tendency of casually mentioning something I've never heard about but absolutely need to have

i really liked LXD, but canonical started to only support it via snap as of late and i really, really don't like snaps. other linux distros didn't seem to be willing to support LXD.

at work i'm moving to docker, my private infra is finally moving to FreeBSD completely (been a long time user for my fileservers already, <3 ZFS)

Oh, LXD. I misread what you wrote. Not so much interest in LXD here.

yeah, it's somewhat of an evolution of LXC i guess?

A framework around it.

i'm not even sure if lxd uses lxc as a backend any more? but that's getting offtopic i guess...

the point was that i'm really happy that jails exist as an alternative :)

Jails are in most ways simpler, although there's something to be said for distinct UID ranges in unprivileged containers.

Until the server goes down and all jails go down with it

mason: Linux containers seem more complex, because they use the same facility (cgroups) for all kinds compartmentalisations

on FreeBSD, network virtualisation is done thru VIMAGE.

I wonder what it'd take to have an include directive for jail configs. I'd like to have that.

one thing we're missing, and which Solaris and Linux do have is process isolation independent of containerisation

solaris crossbow would be great to see in FreeBSD, too

meena: How about mac(9)?

mason: nope, that won't give you what Linux cgroups and contracts give you. mac(9) should be able to act on that facility

I've looked at the idea of a 'null' jail, basically a jail that didn't restrict most things, just created an arbitrary group of the processes inside it

as well as having a separate "namespace" for PIDs inside jails, as well as uid namespacing etc

AllanJude: aaaaand?

if someone has a good commercial use case for it, could totally build it

AllanJude: I have an intermediary: scratchpad.pkgbase.live/Tn-dtb8OT--CttN1Djn8lw#

Title: jrc: Principles - HedgeDoc

I need to understand more about docs.freebsd.org/en/books/handbook/mac

Title: Chapter 17. Mandatory Access Control | FreeBSD Documentation Portal

I've also been a big proponent of UCL

i even started replacing newsyslog.conf etc with UCL, but ran out of steam

AllanJude: a lot of my vapourware is inspired by your ideas

meena: after fixing my config, I can happily report that status and start / restart work with my ergo script :)

reload works, too!

:)

Grabunhold_: yay!

big thank for everybodies time and patience!

Grabunhold_: I'm looking forward to your ports submission.

and if you need help with that, we're here all week. (and also in #bsdports on EFNet)

meena: will probably take a while, i've never done that and i'm in the middle of porting my private irc server from linux / irc unrealircd / a different machine, plus i'm in the middle of moving myself... but it's definitely on my todolist!

i'll also try and contribute the script directly to ergo itself, they're collecting init script examples in their repo

Grabunhold_: i did fuck all during / after my move.

took months to settle.

Speaking of init scripts: I've been using an old style init script for weewx (weather station to web, etc., written in Python) that came in the package. Unfortunately, it doesn't start at boot, but works nicely when manually called as a service.

I've tried to update it according to the FreeBSD docs, but I haven't gotten it to work at all.

See: github.com/weewx/weewx/blob/master/util/init.d/weewx.freebsd

Title: weewx/weewx.freebsd at master · weewx/weewx · GitHub

Well, what do you know. They've updated it. Ill give that a try!

unusual place to put a daemon, but looks like it should work

meena: i'm also overhauling almost all of my IT infrastructure at the same time :D

Seems like it works. I needed to adjust a part or two.

rwp: i thought for a second I hosed a VM

rwp: but, i only uh, did kinda hose it

Why do I get "$daily_news_expire_enable is set but /etc/news.expire doesn't exist" in "daily run output" email? /etc/periodic.conf.local : termbin.com/3rz7

There is no "news" in /etc/rc.conf if that matters

If not for inane messages about things I had not touched, enabled, I would have created/set up /etc/periodic.conf.local often

Oh, and at the end of said email is: /etc/daily.local: No such file