FUZxxl: I ran into your very issue a while back.

FUZxxl: FUZxxl: Did you dedicate the whole disk to zfs or did you specifically setup gpt with a freebsd-zfs partition?

"Basically, next time setup a GPT+ZFS scheme+partition instead of giving ZFS the whole disk. There's not much benefit to the latter and with the former glabel can be used along with sysctl kern.geom.label.gpt=1 variable set."

^ That's what came back, but I'm pretty sure I wrote some more details down about it. I think that section is in my "sort-me-bucket" :/

_xor: no zfs involved

sorry, I mean no gpt involved

FUZxxl: I think there was also something about making sure to do `zpool labelclear` when changing the label, otherwise it would still pick up the old label name (from the copy of zfs metadata at the end of the partition, I think).

the stack is: whole disk -> glabel -> eli -> zfs

ah yeah

I was giving it the whole disk too usually, until I ran into the potential labeling issues.

in my case, what worked was "zpool offline tank label/foo; glabel destroy label/foo; glabel label bar da0; zpool online tank label/foo label/bar"

in the zpool-online step you have to tell zfs that the disk has changed its name

Ah, didn't know zpool-online accept that as an arg.

hm wait

it doesn't

hm hm

maybe it was just zpool online tank label/bar

Yes, I was just about to say lol. I did a man zpool-online and wondered if my man pages were somehow out of date heh.

FUZxxl: Oh, I think what it was is that I would create a GPT scheme with MBR,EFI,SWAP,ZFS partition layout, and then use glabel to actually write the labels (which are persisted as gpt metadata in "automatic" mode).

Just need to make the relevant syctl variables are enabled so that they show up in /dev/{gpt|label}/...

i think you could zpool replace it, but whatever

Great, just great. JRE core dumped when building devel/bazel.

you have more patience than i, to build that

and anything you need to build with bazel

Apparently sysutils/podman-suite requires it :|

Man, why are Google-originated projects such a PITA in the ass to build?

because they use bazel; i just installed the package of podman and it only took a minute

🙄

I'm starting to wonder if there's some explicit internal policy that says something like, "Whenever working on OSS code, make sure to use fragile Linux-isms as much as humanly possible, even where not necessary."

I use an internal repo via poudriere, so I essentially have to build it.

an internal repo that's not git.freebsd.org/ports.git ?

Kind of, that one is the main and then I have two overlay trees, one for custom general ports and another one for private ports for apps I've built.

guys I am almost done with my skeloton project as a skeleton using bazel

Though I was more referring to poudriere:make.conf and options in there, so it could be that devel/bazel requires something that I'm explicitly unsetting, though then that would mean the port is broken in that corner case.

i am going to bazel those arm cortex family

it took me at least two tries to build bezel, but i chalked it up to "running ports over nfs and assumed clocks were off"

What's so great about bazel?

arm: "yamete test me in parallel! i get too fast in development! I can no longer annoy developers!"

My preference is go-task -> just -> cmake/make. I've tried others like scons, but found they were more trouble for me then they were worth.

Evaluating a few implementations of djb's redo though is on my TODO list, just not high priority.

trying to update to 12.3 to 12.4 etcupdate confusing and fucking sucks screw this

After going through every file to to resolve conflicts manually; afterwards it starts merging stuff that wasn't showing as difference when manually resolving and for a single file "Does this look reasonable?" and I press no and I have to go back from the beginning. fucking unacceptable

what does it mean to look reasonable to merge shit I specifically :q out of the file for to avoid being merged

<<<<<< current version is that part of file when resolving conflicts or not? I'm seeing it in vi the editor, this makes no sense

it is much like git conflicts

if you seen such

angry_vincent: maybe i am putting anger on the wrong tool. Is freebsd-update using its own merging method that separate than etcupdate or mergemaster?

i believe, etcupdate is now default, unless you use oler versions of FreeBSD

older*

no

freebsd-update uses neither of these, it does its own diff3

patches welcome

yetoo: yeah, somehow the old mergemaster i thought was easier to use

Like I spend an hour on going through the enter to fix the merge and then after that it goes through the changes I made and it asks if it's reasonable. It looks like I made a mistake and so I press n and then it quits so I run freebsd-update -r 12.4-RELEASE upgrade again and all the changes I made are gone. Who thought this was a good idea? How can I reuse what I spent an hour on? I made a backup of the new and old directories in /var/db/freebsd-update/ is it

as simple as copying them back to the directory but then :q all the subsequent prompts? it's stupid

regarding upgrades.. what's the status of PKGBASE ?

hm can I do path expansion in #!/bin/sh scripts?

GRAYLOG=$(ls /iocell/tags/graylog/root/var/db/{graylog,mongodb}/.zfs/snapshot/daily-$(date -u +%Y-%m-%d))

specifically like that

I'd assumed so, but it appears not to work

dch: I just tried it in 14-CURRENT and /bin/sh doesn't expand {1,2}.txt into 1.txt and 2.txt

yeah, me too

tcsh does

meka: mmmmm I stick to POSIX sh because my brain is small and I am tired

I use fish for day-to-day because it has all the nice things but no need for much default configuration

manpage calls this `pathname expansion` I believe

dch: we're the same on all points :o)

is it safe to share /var/db/freebsd-update between multiple jails (nullfs mount) ?

mage: I'm using it, and i provide a repository: alpha.pkgbase.live

Title: Unofficial FreeBSD pkgbase repository

dammit my jailed zfs datasets don't mount automatically

i wonder if this is a new openzfs CURRENT thing

could someone please update the zrepl port ? :p

mage: why not do it yourself? docs.freebsd.org/en/books/porters-handbook/upgrading :P

Title: Chapter 11. Upgrading a Port | FreeBSD Documentation Portal

rtprio: May I ask you for a favor?

rtprio: Would you please take devel/bazel out back and put it out of its misery? plzkthnx

Morning all

but bazel is blessed by google, it must be good

"good"

I'm having to build bazel because I'm trying to build some newer ports that rely on it. Those ports in turn could end up saving me a bunch of time as it'll preclude me from having to implement my own project to add support for FreeBSD.

I swear though, if it takes me longer to fix/build/whatever bazel then it would to just finish implemeting my project, then...

I'm going to tssk (or whatever it is that British people do) them so hard, they'll drop their tea and crumpets.

if I'm solely in the command line (no gui) is there anyway to copy/paste text on the screen?

Yes, with moused loaded. It's somewhere in the FAQ, IIRC.

Thank you

Title: Frequently Asked Questions for FreeBSD 12.X and 13.X | FreeBSD Documentation Portal

RoyalYorK: if you're mouseless, I'd use copy/paste in tmux.

dch, Run "zfs list -o name,canmount,mounted,mountpoint /jails/myjaildataset" and verify that canmount is yes, noting it inherits from the parent dataset upon creation.

It also would not surprise me if there is a jail.conf to dynamically mount jail datasets but I don't know.

rwp, dch: alpha.pkgbase.live/howto/jails.html mount.fstab

Title: Howto: Setting up Jails

but for zfs stuff, i generally used exec.prestart

And I am not surprised at all that there is a way to dynamically mount datasets for jails. :-) Thanks meena!

Here's a silly question. Can I disallow a user from running `ps` ? Practically I just don't want them to see command line args of a process

it's for a CTF ^

hello again

antranigv, IIRC it is the sysctl security.bsd.see_other_uids=0 and security.bsd.see_other_gids=0 settings.

rwp that I do know, I was wondering for the same user :D what could go wrong if I delete `ps` from a jail?

CTF? Okay. Deleting the executable would not disable the kernel syscalls though. If anyone can upload a new binary then that would not prevent them.

Also if it is in a jail then deleting ps really can't become too terribly bad, for a CTF it would not stop me from doing it, but some startup and shutdown scripts call ps to operate and will then fail if it is not present.

rwp: I always wondered why we don't use pgrep in rc

meena, I can only guess that since pgrep is a "new" addition in the grand scheme of things that it wasn't always in the base core but ps has always been.

The man pgrep page says pgrep first appeared in FreeBSD 5.3 modeled after the utility from other systems.

_xor: lol

yeah, rc is a lot older than 5.3

and switching FreeBSD rc to use pgrep wouldn't really do any harm wrt merging anything from NetBSD, because the code diverged already long ago

!!!

meena, absolutely fucking no.

When you start a process, you write a .pid file.

Then you use it to stop.

pgrep is some ghetto ass shit.

CrtxReavr: are u OK?

Not with that kinda nonsense flying about.

yeah, chill out CrtxReavr

racial-slur-adjacent?

"ghetto ass shit"

That's in your own mind.

In any case ftp-proxy is the only thing I immediately find which uses ps for pids.

sendmail uses ps to generate hard to guess random seeds for certificate use.

antranigv: what is CTF in that context?

That's different; that's for entopy generation.

Though, there's still probably better ways to do so.

I am sure it is Capture The Flag. A computer security competition.

yuripv, en.wikipedia.org/wiki/Capture_the_flag_(cybersecurity)

Title: Capture the flag (cybersecurity) - Wikipedia

grep entropy /etc/defaults/rc.conf

rwp: i see, thanks

i'd still love to know why pgrep is ghetto-ass-shit. I wonder if other Solaris inspired Software is also ghetto-ass-shit?

it's about usage; you want to stop process that you started, not something you randomly match

in that case, we should port contracts from Solaris, because pidfiles are a plague

The problem is really about name collisions. Say you want to run multiple different nginx pools with different configurations. Then later want to stop/reload them. They don't identify by name well as they are all "nginx" by name.

contracts are really cool…

then you'll want smf to go with it, and then all out fighting begins

basically like Linux cgroups, but less awful

rtprio: Got it to build, weird reason why it was failing. java/openjdk11 was building the previous minor version, even though I updated the tree to head.

Speaking of all out fighting remember we got here because of a Capture The Flag competition question, WCGW deleting ps from a jail?

Oh well, whatever. Manual reset and update seemed to do the job.

feast your eyes on the vaporware occupying my brain for years now, scratchpad.pkgbase.live/Tn-dtb8OT--CttN1Djn8lw#

Title: jrc: Principles - HedgeDoc

at least you are not proposing 4th

yuripv it's a Capture The Flag competition! it's a hacking contest :) I've been deploying them on FreeBSD for years!

yuripv, You don't like RPN calculators? Heresy! It's the one true way.

meena, I am still cheering for runit, s6, nosh, or one of those variants.

antranigv: there is also pargs then

also top and whatever other utilities could show your the args

rwp: looking at the loader's .4th makes me shiver, i didn't look further than that :D

I admit I haven't spent any significant time with fourth. But at least it has an interesting language paradigm. But lua does not excite me.

meena: "Fault Manager, however, is Golden Grail of these designs", having recently been debugging mysterious memory leaks (which are not really leaks after all) in fmd, i would say that it's monstrosity, combining a lot of moving parts in it :)

i don't think anything would go wrong deleting ps

Out of curiosity, do you any of you use the **PURE** cli (no gui installed) or do you use the gui and use xterm or something along those lines

I only use X Windows (the gui) on my desktop. All other systems not my desktop being used as servers without X.

what is pure cli? i ssh into my servers.

I should have been more clear. What I meant by pure cli, is that when you boot up the system are you going directly to the tty and staying the tty or are you booting into a gui and using a terminal window

if you ssh, then you would be using the tty

Oooooh very niiiice.

OCI image looks to be running :D

I have managed to get xfce working from the tty (slim is another story), but I have more to learn in the terminal first before I use the gui on a regular basis

gui with multiple xterms is the way

i dont have the visual memory to flop between consoles so well, much easier to compare side by side

yuripv: unfortunately, even golden grails in software are still mostly made of software.

Check out wezterm, it's my default terminal these days. It's so amaze.

Been using it for a while now and it's the best terminal I've used thus far.

i donno, i don't think i could use the cli if not for `tmux`

I was a heavy tmux user before, and I still like it, but don't have as much need for it these days with wezterm.

i'm using terminology nowadays

i'll check out wezterm

rtprio: Wellp, looks like I won't have to finish writing my task driver after all.

does nexus work better on freebsd now?

Though I think I still will since I've already implemented a non-trivial chunk of it.

rtprio: You mean the artifacts repo by SonaType?

oh;

Yeah, I got those new ports I mentioned earlier to build and just now set them up to run OCI images with my orchestrator.

I was willing to spend up to an hour fiddling with it, but it only ended up taking 5 minutes.

hi all, I just updated two of my freebsd boxes, 12.3p10 -> 12.4p0

do I need to do something for pkg also?

rubin55, Yes. Also run "pkg upgrade" to upgrade the binary packages.

And then reboot. With that upgrade I found that after the "freebsd-upgrade install" that sshd needed to be restarted, which of course happens on reboot.

Release notes for that upgrade freebsd.org/releases/12.4R/relnotes

Title: FreeBSD 12.4-RELEASE Release Notes | The FreeBSD Project

mm, yes, I did that, but got no new updates.. since it's still 12.x, I didn't think too much of it, but wasn't sure

Oh, I almost forgot to mention "pkg autoremove" to clean up no unneeded packages.

If you have already upgraded all of the binary packages then I think all is good.

On a major OS release such as from 12 to 13 it has been recommended to reinstall all of the binary packages "pkg upgrade -f" to ensure that all shared libraries match.

right, fair enough!

The Handbook section on upgrading: docs.freebsd.org/en/books/handbook/…e/#updating-upgrading-freebsdupdate

Title: Chapter 25. Updating and Upgrading FreeBSD | FreeBSD Documentation Portal

That's a little confusing to read because it also covers source compiled upgrades intermixed in there.

But basically you have done everything you need to do for this upgrade from 12.3 to 12.4 so Good Job!

6 years! from 10.x to 11.x to 12.x

I'll note again the caution that on that for me 12.3 to 12.4 required sshd to be restarted before it would accept new logins.

Congratulations! You are an old-timer now for sure! :-)

every time I read freebsd docs I'm so impressed..

lol

rtfm actually means something in bsd land

One thing I really very much appreciate about FreeBSD is that the base system is one cohesive whole. This makes upgrading reliable and safe.

rwp: this so much

@rubin55 yeah it means read the friendly manual instead of read the f'd-up manual.

Quantafac: exactly :)

hey, how do I list attached USB devices?

Network_Jack, If you are looking to see USB Storage then try "gpart show".

not storage, I am looking for a USB network device

Network_Jack: dmesg should list it (or /var/log/messages when attaching/detaching).

`usbconfig` shows all, that is attached to USB

Network_Jack, Try "pciconf -l" and "pciconf -lev"

Kalten, thaanks

Oh, sorry, I missed the usb entirely with my pciconf commands. Ignore me.

you might have to load a kernel module or it could be usb-generic.

And I think, `/etc/devd.conf` (and the `/etc/devd/` directory) could be used, to automatically react to attaching and detaching.