it is not a usb dongle

it is an actual card!

amazing

Title: PCE-AX3000｜Adapters｜ASUS Australia

Normally you find them on Mini-PCIe or M.2 - I have an AX200 which uses the latter interface.

Intel even make desktop kits where you get a AX200 and a PCIe to M.2 daughterboard.

Mini-PCIe?

you mean, you can get a mini mini?

Mini-PCIe is the interface that was used in (most) laptops before M.2 was introduced.

I bought it probably four years ago

I have M.2 slot, but it has a SSD in it

M.2 added keying so you could connect both PCIe and SATA devices.

U.3 (which I'm hoping will be the next interface), lets you do PCIe, SATA, and SAS over the same connector.

oh well

I cannot get the bluetooth to work

presumably because I do not know how to get firmware to load for pci-e only usb with iwmbtfw

Yea, I don't believe bluetooth works for iwlwifi yet.

iwlwifi(4)

back to a dongle?

At some point I got a handful of USB Ethernet devices that use AX88179, which is a chip that's supported by everything.

It's a direct USB3 to Gigabit Ethernet PHY/MAC so it can do wirespeed (as USB XHCI is packetized anyway).

meena: maybe it was, but i still missed it :D

dunno

There's also a USB-C 3.1 Gen1 to Gigabit Ethernet PHY/MAC from the same company, but I forget the model number.

I have a bluetooth dongle somewhere

not as reliable tho as the pci-e card

That one does CDC-NCM so uses cdce(4).

Oops, USB-C 3.2 Gen1*

Now if ASIX (the company behind them) was _really_ smart, they'd do a USB4PD to SFP+ connector for 10GbaseT.

though I sometimes used FreeBSD UNIX since 1997 (yes, it was actually called UNIX in official documentation then and is still a 'genetic UNIX') I forgot how to see extended partitions in fdisk... I had more experience with DOS & GNU/Linux fdisks...

trying to mount EXT4/Android/BlissOS from some extended partition

p

okay, I should use gpart instead. The drive appears to be nvme0 but 'gpart show nvme0' says 'no such geom'

oh, nvd0

says same for that even though it's what's used in fstab

and mount shows that '/' is currently on that drive

also, how can I check /etc/ttys to make sure (other than me adding more ttys) it's up-to-date?

i was still able to add that to fstab and mount it, but gpart won't show it

it does have a ZFS on one partition which I saw people having problems with 'gpart show'

finally found out it's nda0; so which is correct for fstab: nda0, nvd0, nvme0? Do fstab & gpart really use different defaults?

even with nda0 I still can't show the extended partition

i thought *BSD was still planned/designed/standardized ahead of time, not 'make up as you go along' two instances of the same differently like GNU/Linux does?

Does anyone have a document that explain show to install OpenGrok on FreeBSD?

I can't seem to figure out from google...

happy new year, bitches.

pretty sure...the opengrok port is broken...

I'm a bit surprised my /root/.shrc file isn't executed when I su to root, what am I missing?

are you on 14.0-release? and did you change the relevant line when upgrading?

crb: "su" or "su -"?

su and su -m, it looks like su -l will invoke it

Yeah "-" is same as "-l"

pkubaj, hi! scribus.net/scribus-1-6-0-released

Title: Scribus 1.6.0 Released – Scribus

eee

ansible users? I'm trying to use a dictionary to populate a bunch of configuration files - However, I'm getting "with_dict expects a dict" - I don't see why the supplied parameter is not a dictionary. gist.github.com/dlangille/13851af2f8cb6bf028d2ca43e42b5d48

Title: with_dict expects a dict - why is it not a dict? · GitHub

Solution: use loop: "{{ proxy_websites|dict2items }}" or with_dict: "{{ proxy_websites }}"

VVD: thanks, but the port i maintain is scribus-devel, which implies a development version

now i'm wondering what to do here

one thing would be to rename it to just scribus, now that the stable scribus port was removed

another would be to just upgrade to 1.7.0 once it's out

well, scribus-devel seems to be at 1.5.8, so still a reason to update that

pkubaj, you can add scribus with 1.6 and keep scribus-devel for 1.7 branch

anyway we need 1.6 and probably 1.7 in future too

bsd neweb here, how can I grant root access to my new box normal user, while freebsd prevents me from using su - root or su -l root

"In particular, by default only users in the “wheel” group can switch to UID 0 (“root”)." (from man su)

any concrete command to try?

add the user to the wheel group using pw

vigr is also quite nice, if one can control vi

it respects EDITOR according to its man page

oh, cool, didn't know that since vi is my editor :)

pw groupmod wheel -m <user>

so that as root and you will have that user added to the wheel group, which will allow su

$ pw group mod wheel -m "username" >> pw: gr_tmp(): Permission denied

librton: you need to be root to do that (so log out and log in again as root)

pw user add "username" -G wheel >> pw: you must be root

unixwitch: that was the main question, how can I become a root when su - root and su -l root are being denied?

librton: log out, at the "login:" prompt, type "root" then type your root password. once you've run this command, you will be able to use su instead

oh ok

ugh, not able to do so, I'm connected via ssh and I don't have a keyboard at the moment ;(

in that case you are out of luck unfortunately, you can't log in as root over ssh by default, so you need to log in on the console to add your normal user to wheel

ssh not configured to do root access

it used to be you could run /usr/bin/login to do this, but that is not setuid anymore, so it doesn't work

librton: If FreeBSD is installed on a physical hardware or in a VM, access the machine physically. If it's on a VPS, there's a good chance your VPS provider will offer you a web-based console, from where you will be able to login as root.

unixwitch: doesn't that defeat the purpose of the wheel group

Hello71: yeah honestly i don't remember how this worked. i think i only ever used it on systems that didn't restrict su by default anyway...

remiliascarlet: physical box, no physical keyboard

Plug one in then.

Unless that physical box is from the 1990s or earlier, it'll have a USB port. And USB keyboards can be bought for dirt cheap.

presumably a keyboard of some sort was originally attached to install the OS anyway

mildly annoying but well, BSD has worked like this forever, so it's just something you learn :-)

Unless he did so through telepathy.

remiliascarlet: i was thinking maybe PXE but it seems unlikely a newbie would install that way

updating to python 3.11, wish me luck... [main-default] [2024-01-01_17h19m24s] [balancing_pool] Queued: 291 Built: 0 Failed: 0 Skipped: 0 Ignored: 0 Fetched: 0 Tobuild: 291 Time: 00:00:08

I've been using Unix-like systems exclusively since the early 1990s, and I still can't understand how PXE works, left alone a newbie.

i did a lot of PXE installs of Solaris, but they included a whole bunch of scripts that basically set everything up for you automatically. never tried it with freebsd

i did PXE install Windows once, that was actually much easier than i thought it would be. it can even join a domain during install

(i can't even remember why i did that, i think i lost my go-to 'install a new OS' usb stick...)

... ok bye then

people of freebsd, hello ;/

what's format of .sh_history that I see some "\12" and similar numbered codes in every history line there?

is there a way to get rid of that nonsense and have just histor without that?

or do I have setup something wrong that I have that in histfile?

alright guys I've figured it out, very simply put used "sudo su -" landed me into root rightaway

hura

next command is how to update upgrade freebsd package manager

librton: sudo isn't installed by default, is that something you added yourself? ... for updating packages, use "pkg update" then "pkg upgrade"

unixwitch: it was a month or so ago, I can recall .. I must have fiddled a little "when I had a keyboard"

(note that pkg upgrade doesn't restart services, so you may run to either reboot or manually restart things after running it)

s/run to/want to

No updates needed to update system to 13.1-RELEASE-p9.

that's the output from freebsd-update, do you have a question about that?

It would be useful background information for the rest of us to know how this system came to be a bare metal remote machine with FreeBSD installed to this point. It seems that someone else installed it for you and set you up on it? And they partially set it up already?

morena: from what I remember it's a custom format which libedit uses to save history to files and read it from them. For plain-text history I think you'll have to use fc -l

First make sure you are in the wheel group so that your account can upgrade privilege to root later.

To upgrade the base system "freebsd-update fetch" followed by "freebsd-update install" then reboot then "freebsd-update install" again (if needed) then "pkg upgrade".

oh, hi pstef. (you commented on my bintrans PR on github)

morena, \012 is octal and the ASCII code for an LF line feed which is the Unix line ending character that is read when one presses Enter.

unixwitch: hi :)

morena, \040 is octal and the ASCII code for an space character.

rwp: ye, but it should be and it is unavoidable in histfile?

I never saw that in any OS before ;/

it's what libedit does, sorry

okay, so at least nothing wrong at my side

still interesting that freebsd went this path ;/

Christos did, we only import his library to handle history in /bin/sh

oh okay

you did not check before how other handle it?

the choice to use libedit was before my involvement with FreeBSD

sad, you came late

but at least sh is usable and not too raw ;/

"Please click on all images of the heaviest animal species" these captchas get more bizarre every day

you can always install another shell from the ports tree

rather not, there will be more headache ;/

sh feels like it's enough for my simple needs

It's worthwhile to install what's comfortable.

i think the only options here are libreadline (GPL), libedit, and libtesla, and for whatever reason libedit was chosen to be imported. possibly tesla didn't exist at the time

Kinda like spending the money on a good keyboard, 'cept the software is free.

(and obviously libreadline is not an option because of the license)

i actually can't remember if libedit was important, or if it originated in freebsd and was exported after, like libarchive

s/important/imported

will see if something will not start write binary into my home soon ;/

morena: it's not *that* hard to parse some \escapes really, is it?

and still you have fc -l

no it's not, luckily other software does not write that kind of nonsense into config files na similar ;/

think of it as a database (of history entries) rather than a config file

ye, there are many solution obviously, it just feels to me wrong to be there

at least it's not sqlite :-)

oh ye ;/

sqlite would have made it possible to store multiple sh sessions properly

unixwitch please don't encourage him

pstef: but it would break if /home is on a 9p filesystem since sqlite really likes mmap

well, i guess freebsd doesn't have a 9p client anyway. but if it did...

I could only find reviews.freebsd.org/D41844

Title: ⚙ D41844 Add an implementation of the 9P filesystem

oh... is this in current? i should investigate this

ok no it's not, i guess it was reviewed but not actually merged yet

this would be really nice for sharing files with bhyve guests though

I believe that's the primary motivation for this diff

dfr's is going to be the most up-to-date version, even

well i subscribed, once it gets merged i will probably be motivated to work out what's broken multimedia/jellyfin so i can move my media server vm to freebsd

hi. why this not works? see paste: ctxt.io/2/AADQWwA4FQ

Title: Context – share whatever you see with others in seconds

gorkarabela has joined

Z

zagorkarabela 21:27:47

hi. why this not works? see paste: ctxt.io/2/AADQWwA4FQ

button event works but axis event not works. cursor cant move.

have synaptic toucpad and try this on pure X11.

Title: Context – share whatever you see with others in seconds

atkbd keyboards works btw.

generic kernel btw

happy new yer btw

on linux it sends absolute axis events. I am using Xorg server nor wayland

freebsd handbook says Xorg auto{detech/configure}  input devices no need for manual xorg.conf

this no config approach use catch-all semantic under /dev/input and seen unrelated devices as input devices say lid swich. So i have to resort manual xorg.conf

Is linuxkpi/mesa-dri  mandatory for regular aps(not games) in xorg-server case? I am try to keep away dependency hell So have a look to xorg to see whats the stuation.

say firefox, pcmanfm

meena: do you know the right person to annoy to get my ports PRs landed?

unixwitch, which PRs would that be?

unixwitch, if you have maintainer-approval+ or timeout I can help.

jbo: 275944, 275945, 275947, 275959, 275962. not maintainer-timeout, these are new ports

unixwitch, I can grab them, but I am still under mentorship so I need to get an extra round of approval from my maintainers before I can push the commits - does that suite you or do you need an ASAP solution?

jbo: ASAP is not required at all, it's just (mildly) annoying me that i need to update them by hand after a poudriere run :-)

jbo: as no one else seems to be picking them up i would be very grafeful if you did

morena, To me this feels not like an OS thing but like an sh implementation thing. All of the shells save history in their own format. bash, ksh, zsh, all have unique history formats. sh having a unique history format seems like an sh thing not an OS thing. Just like the other shells.

unixwitch, done

unixwitch, expect little progress until next week starts. stuff movies slowly at the beginning of the year.

jbo: thanks!

yeah no worries

I'll do my part of the work the next few days. probably gonna take two to three days based on my current mentor's load.

morena, I don't know why anyone would care how the sh saves its history format because I think "history" or "fc" should be used to read it anyway.

I guess I have "history" set as an alias to "fc -l".

jbo: note (if you didn't already see) some of these have interdependencies, mainly audio/sublime-music (#275959) depends on the other py-* ports

unixwitch: if i don't have someone like jbo jumping in, i join #bsdports on EFnet, and/or #freebsd-ports and post my (or other #freebsd-bugs ) issues there

unixwitch, I still feel like bugzilla is having non-intuitive ways of labling blocks/depends-on :D

unrelatedly, i think i'm misunderstanding something about ports or poudriere. i set DEFAULT_VERSIONS+=python=3.11, did a poudriere bulk, it rebuilt ~300 packages... but it didn't build a python311 package?

#freebsd-ports +1

jbo: i like the tree view bugs.freebsd.org/bugzilla/showdepen…ytree.cgi?id=275959&hide_resolved=1

Title: Dependency tree for Bug 275959

275959 – [NEW PORT] audio/sublime-music: GTK3 desktop client for Subsonic-compatible music streaming servers bugs.freebsd.org/bugzilla/show_bug.cgi?id=275959

meena: this reminds me, i mean to ask if #bsdcode is still around, it's been so long i can't remember the key

key?

the key (password) to join the channel

AFAIK, #bsdcode on EFnet has never had a key.

unixwitch: it's around, open, and very dead

rwp: no idea, somehow after decades I first time see some software throwed at my home some hexa or whatever codes into plain text ascii file

hmm, really? maybe i'm misremembering, i'm sure it used to have a key

(this was back in like freebsd 3.x days)

rwp: nothing critical, just first time experienced, so was not sure as online search did not tell much

Maybe in the 3.x days. . . . was prolly the 4.x or 5.x that I first had any reason to be in there.

i remember someone (does anyone remember who?) talking about the super-secret new "HAMMER" platform they were porting the kernel to... which nowadays we call amd64

s/into/Intel

i'm conflicted about that, i thought Itanium was nicer for getting rid of the legacy stuff, but it was basically a terrible platform because of how difficult it turned out to be to write the compiler

ARM is strongly predicted to degrade Intel further.

they fixed that for Itanium 2, but by then everyone was using amd64 and it was too late...

Itanic was. . . very problematic and slow.

all else being equal, i think we'd be in a better place today if people had adopted Itanium (2) rather than amd64, but it's pretty obvious why that didn't happen

I had a rack full of prototype Dell IA64 servers.

Well, $job did, but I was the only one interested in playing with them.

We were a huge Dell shop.

i was fairly involved in OpenVMS at the time so i remember all the HP systems that used it

HP basically sustained Itanium for 20 years on life-support because of their investment into it

At that time, the only HP systems I messed with were PA-RISC.

And not in large numbers.

Itanium was supposed to replace PA-RISC and Alpha and be a unified platform for all their systems

We had a CTO who was the project manager for the Alpha at DEC. . . he had some interesting perspective on it.

PA-RISC was fun too though, it had no atomics at all except for a 'set 0/1 at memory address' so all the atomics used an array of pthread_mutexes

Everyone blamed Compaq for Alpha's demise, but DEC knew before they sold it, that Compaq was buying a moldy turd.

you'd index the array by address of the variable and lock it to do an 'atomic' update...

CrtxReavr: everyone hated the HPaq merged but in hindsight, it probably saved HP, since nowadays all they sell is x86 servers

s/merged/merger

Well. . . Compaq sold solid, reliable (perhaps pricy & over-engineered) x86 machines. . . whereas HP sold utter shit.

(... which are just rebranded Compaq servers, they abandoned their own product line)

no clue why my vlepy.com wont respond to icmp

HP could never escape their ~20% DOA rate.

Even on top of the line, HP 9000 PA-RISC servers.

mane: we have no clue either, you probably need to provide more information. like, is that system running a firewall?

pf

show pf.conf and also 'ifconfig -a'

unixwitch paste.debian.net/1302705

Title: debian Pastezone

unixwitch and this is ifconfig paste.debian.net/1302706

Title: debian Pastezone

mane: well you don't have a pass in rule for icmp which probably explains it

wasn't Itanium the… spiritual successor of the Alpha?

meena, not sure what would make you think that.

meena: err... maybe? it was a completely different design and had almost nothing in common with Alpha aside from being RISC...

it was the successor in the sense that DEC (after they became HP) switched to Itanium but that was more of a marketing concern

unixwitch ok got it

ah

DEC knew from the start that the Alpha was a design fluke, and that they'd never be able to make it keep pace with Moore's law.

mane:

PING vlepy.com (176.56.237.86): 56 data bytes

64 bytes from 176.56.237.86: icmp_seq=0 ttl=53 time=56.880 ms

works now

yeah i just set a in rule in pf

thanks

i switched to fbsd after 10 years of using debian :DD

mane, also. . . the SMTP(s) server in your MX record doesn't connect - just times out.

ok

morena, By my thinking sh even having a history just feels strange. It never had a history file before. So whatever it has now is new and how it does it has no legacy compatibility issues.

rwp: i assume this was part of the "make /bin/sh not suck" project which lead to changing root's shell to /bin/sh

csh had a history file. ksh had a history file. bash has a history file. zsh has a history file. But sh never had one before. I realize people like it but then since it is new there is no reason for it to be any particular format.

there wasn't a project, a few people felt annoyed with the default shell but at the same time refused to switch to tcsh, me included

I set my user to bash and 'sudo -s' lets me keep my shell, so. . . win win.

unixwitch, My problem with "make sh not suck" is that it is still not as good as other shells by a lot so it's going to be too little too late for me regardless of what is there.

so trasz did some prompt stuff, bapt enabled history, I wrote program name completion, etc.

pstef: 'project' might be the wrong word, but weren't there a bunch of commits made in order to justify the change?

bapt also made other changes beside history, like emacs bindings

rwp: sure but how often do you actually log in as root? i'm just happy that 2>&1 works and i don't have to (re)learn csh which i hate whenever i log in as root

eww

I thought all emacs users were required to use zsh and OhMyZsh.

Re sh, more importantly vi-bindings!

The trifecta. .

it's not about emacs users, for example I'm not one

I am always doing things so probably I log in as root more often than most people. I set up toor for my own use leaving root defaulting to only things in base.

i am always doing things but i use sudo -s...

but ctrl+left and ctrl+right jumping between words is pretty much the standard everywhere except places that want to be original

CrtxReavr, vim + zsh user here :D

wait, aren't vi and emacs bindings required by posix for /bin/sh?

/bin/sh is not POSIX

nimaje, what about OhMyZsh?

See: local, pipefail

parv, sure it is.

CrtxReavr: why would I look at that shit?

1975 called. They want their shell back.

(i mean... everyone uses tcsh nowadays anyway right? so there's no reason to ship /bin/csh)

csh was the ultimate paper printing terminal and mostly became obsolete when the CRT terminal arrived.

unixwitch: it's been moved to their own PkgBase package, so if you're using PkgBase you can just not install it

For the last twenty years csh has been a flavor of tcsh source.

meena: i look forward to stable pkgbase appearing in FreeBSD 24.1

138847 -r-xr-xr-x 2 root wheel 425456 Aug 2 21:12 /bin/csh

138847 -r-xr-xr-x 2 root wheel 425456 Aug 2 21:12 /bin/tcsh

unixwitch, ;->

There is no separate standalone csh that is not also tcsh.

unixwitch: I'm hoping for 16 or 19

They may behave slightly differently, depending on which name they're invoked as.

Not sure I would notice though.

nah

There are still some very large tech companies (I am looking at you Intel) where the lab still is mostly csh throughout. It's crazy.

tbh i don't even want pkgbase because i am always doing 'cd /src/main/usr.bin/foo; make install' and i guess pkg doesn't like that

that's not been the case for years now

AT&T banned all bash and bourne shell usage, because of one quickly patched bug.

you can't use csh for scripting tho…

you can you just shouldn't

You _can_ use csh for scripting, but you never should.

Well, you can. . . if you like really shitty scripts.

you can't even have functions

:D

I just killed three ancient csh scripts rewriting them in portable shell for a client just two weeks ago! They are still out there. Hiding in the weeds.

i have no sympathy for csh users, i learnt Unix in the 90s and even then everyone said not to use csh

rwp++

the client?

One of my clients with a long legacy installation.

I wanted to say that I am an emacs user and don't use zsh nor do I even like it, zsh is just too weird for me. But it is okay with me if other people like it. Just not me.

at least zsh is "POSIX compatible"

when we only had /bin/sh or zsh, zsh was a shit. i understand people using bash nowadays though, it's improved a lot

csh is just no.

s/a shit/the shit/

to be honest I never heard somebody use csh

zsh was like, it had all of csh's feature, but it didn't suck

so why would you use csh instead of zsh?

I use bash but honestly I don't like a lot of things about bash and I never write bash scripts, I write portable shell scripts, but I haven't found a shell I like better for the command line. I dream of eventually writing my own so it would be the way I like things to be.

NeXT even made it the default shell, iirc

the FreeBSD guide used tcsh syntax in a few places, but sh syntax in all others

History recall of multi-line command is atrocious in bash 4.x compared to zsh, where each line gets its own history instead of one-complete block

When /bin/sh was the original shell it made sense to use csh for root's shell so everything was in base.

i've been trying shells/fish recently which i have to say i'm actually quite enjoying so far

... at least in vi-mode.

rwp: it really didn't. System V kept root's shell as /bin/sh. BSD only made it /bin/csh because they were precious about /bin/csh being a special 'better' BSD thing

... if not for that, I would still be using bash instead of zsh

I actually use the multi-line conversion to single line behavior a few times to turn a multi-line script into a command line one-liner. (shrug)

unixwitch, I meant in FreeBSD. Right. I spent years on HP-UX at work and it is System V and always had root's shell either /bin/sh or later /bin/ksh

just until recently, FreeBSD sh would print an actual tab when you hit the tab key, and not print a newline when you hit ^D

yeah ksh was alright

did freebsd ever get any ksh in base?

I blame Sun for some badness though because their /bin/sh was the old Bourne shell and the POSIX shell the put in /usr/xpg4/bin/sh which is the reason #!/bin/sh is not written into the standards and that is going to haunt us forever when it comes to portability.

guess not fish: Unknown command: ksh

20:00, <unixwitch> so why would you use csh instead of zsh? -> I always grown up with the warrent that it did not matter if it was (t)csh or sh but the indeed as long as root always was a base system shell

so you could login no matter what

toor from the older years was always bound to /bin/sh for that very reason I believe

unixwitch: i love fish. mostly because I'm dyslexic* and also, because sh derivates are painfully inconsistent, and my brain can't keep remembering all that garbage

With other systems going through UsrMerge right now I am seeing all kinds of crazy things such as "#!/usr/bin/sh" (gack!) and now the crazy "#!/usr/bin/env sh" is feeling less crazy now.

daemon: yeah i agree with that. i meant more for a user shell. if you login in as root and have to type 'zsh' to get a nice shell that's fine

daemon: i just don't see why anyone would keep using (t)csh after zsh existed.

I just discovered that current sh only stores 100 items in its command history by default, you have to specifically set HISTSIZE. And that our new history alias should have been 'fc -l 0'

zsh for tmux for the god combo but yeah

meena, Does fish not deviate how it works wrt other shells?

s/for/plus/

parv: it's not a POSIX-compliant /bin/sh if that's what you mean. but it doesn't claim to be

parv: posix doesn't disallow extensions, but yes our sh isn't posix (see '-c' bug 274650), sometimes it's better to break portability than have users produce broken code by accident (well, still easy in sh, just use zsh)

274650 – sh does not accept -- after -c bugs.freebsd.org/bugzilla/show_bug.cgi?id=274650

Title: fish shell

I still quite like tcsh's default history completuin

completion*

"Finally, a command line shell for the 90s"

the only thing i don't like about fish is i can't work out what the equivalent to !* is

this is probably buried in the docs somewhere...

what is !*

$ badecho foo bad baz

zsh: command not found: badecho

$ echo !*

echo foo bad baz

foo bad baz

it's the arguments from the previous history entry without the command itself

ah, no idea

it also doesn't seem to implement ^foo^bar history substitution but i find i don't use that very much nowdays so i can live with it

could easily hack a quick command in perl to replace it

daemon: i am mildly disturbed about this suggestion

its what it was made for ;)

i was so happy when perl was removed from base

no happy memories about that at all

if you have zsh, tmux, nginx etc; you very likely have it anyway

i do have it but i dream of one day submitting enough patches that i don't

openssl reuires it to build still ;)

that isn't much of a recommendation :-D

does libressl need perl?

good question not actually sure lets see

ah, shit, why would xorg be removed if I removed perl? even xorg-server and xorg-drivers, not just the metapackage

meena: omg fish doesn't have an irc channel it has a matrix channel. this is so terrible

perl is involved in a lot of build or make files

could be worse; could be slack

nimaje: turns out perl is fricking everywhere bsd.to/MNaR

Title: dpaste/MNaR (Plain Code)

like a virus

;

Title: 224: Lisp - explain xkcd

;)

wait, why do i even have xf86-video-scfb installed?

219 packages that would get removed for me

freshports.org/security/libressl no references to perl directly

Title: FreshPorts -- security/libressl: Free version of the SSL/TLS protocol forked from OpenSSL

There had been efforts to replace Perl with awk+sed in The Ports. Someone(s) offended enough just need to extend that to other projects.

unixwitch: because of xorg metapackage?

nimaje: yeah but i thought i changed the options. i think this was an orphan...

pkg autoremove ?

i dislike pkg autoremove because (ime) it keeps trying to remove things like ca_root_nss

then mark it as manual via pkg set

i know that, but i still don't trust it because if it removed that what else will it remove?

although actually this reminds me of something i wanted to ask. what X.org packages do you need for Wayland? it seems like Wayland compositors still depend on X keyboard layouts, for example...

"pkg-autremove" prints the list of things to be removed & prompts. Unless one mindlessly says "yes", as I once did.

i know. my point was i don't want to blindly accept what pkg autoremove tells me to remove.

stuff you installed via tranistive dependencies and throught you installed manually, but it will show you what it would remove and ask you if it should continue with that, so check that again and mark stuff you missed as manually

it's also awkward because i have some packages installed via 'make install' from ports i maintain which haven't been merged yet

hm, but these packages should be properly registered in the database or do you mean build dependencies?

hello

nimaje: yeah but they still show up in pkg autoremove if they're build depends

unixwitch: if pkg autoremove is removing things it shouldn't, you should submit it as bug

well i did a pkg remove xorg, pkg autoremove, and restarted Wayland and everything seems to be working

meena: it's probably not a bug, but i found it odd at least in the case of ca_nss_root since i thought curl depended on that

ca_root_nss rather

curl doesn't depend on ca_root_nss, cgit.freebsd.org/ports/commit/?id=4…ceec92a58dae0979a2744fe71fca8675cf9 fixed that

Title: ports - FreeBSD ports tree

i mailed the person listed as MODULE_AUTHOR in the Linux driver for alc(4) and they replied to say "I have neither documentation nor any knowledge about these chips, if I made any modifications to their driver it was incidental to wireless stack work." which alright, understandable, but then why are they listed as the author?!

indeed

what does git history say?

meena: do you want really me to clone linux.git, this is going to take months

although it's not a bad suggestion actually

so maybe i'll try that

nah

just go on Github, and press the blame button

oh clever

lazy

i'm 100% sure i will fibre to my desktop and switch alc(4) for ixgbe(4) before i actually fix this alc bug

if it's something i can do on my phone, I'm not getting up for, and I'm certainly not downloading six gigabytes of history for one file

i might just submit a PR to disable MSI-X on Killer Ethernet by default on the basis that working slower is better than not working at all for new users

I have to be pedantic. . .

It's only spelled "fibre" in the case of "fibre channel."

CrtxReavr: i am british, we spell it 'fibre' here

fibreboard, fibre ethernet, fibre channel

All other forms of fiber optic cable are "fiber," Britishisms be damned.

CrtxReavr: also, how do you know that i don't intend to run Fibre Channel to my desktop and use IPoFC?

i mean, i'm not going to do this because it would be silly. but *you don't know*!

I've not heard of IPoFC, though I'm sure it exits somewhere. . .

i actually don't think freebsd has a driver for this. Solaris users liked it

However, I very much doubt it has any useful utility.

probably because 8-16Gb/s FC was cheaper than 10Gbps Ethernet at one point

ip over fiberchannel?

yes

I have a very hard time believing that.

Since I worked for EMC when 10 Gbit ethernet was coming of age, and we were also a huge Solaris shop.

Title: Configuring IPFC SAN Devices - Oracle Solaris Administration: SAN Configuration and Multipathing

is FC lower latency than ethernet?

And we tried and added NAS and SAN storage protocols, like it was going out of style.

meena: i am not an FC expert but i would imagine not. FC switches use the same basic technology as Ethernet switches.

I doubt it. . since ethernet uses six byte hardware addresses and FC's WWINs are. . . huge in comparison. . . and they're both very low-level protocols.

FC itself was sort of a layer 2 protocol but IP is L3 and relies on the L2 transport by definition...

i'm pretty sure the reason was that back when servers came with 1Gbps Ethernet as standard, they often included 4/8/16Gbps FC HBAs and some people wanted to use those

When did Oracle acquire Sun?

err, 2004? ish?

more like 7 or 8

oh, 2010, according to Wikipedia. wow, was it that late?

announced 2009, finalised 2010

I was never a solaris fan, though I had to support it a lot. . . who'd have thought Oracle would make it worse.

(Don't answer that.)

during my first job, which was Solaris

We didn't like the result

i only use FreeBSD because Oracle killed OpenSolaris

hahahaha

if they hadn't i'd still be using it, it's the most perfect Unix in my opinion

ooooof

I was most pissed when Oracle acquired Xsigo.

I came to FreeBSD via the ASF, but stuck to it thanks to Solaris

That was awesome technology. . . and Oracle fucked it all up.

meena: one day i am going to write dladm(1m) for freebsd and submit it as a PR

imagine being able to configure network interfaces and having the live config *and* the boot config updated at the same time

inconcievable, right? but dladm did this!

illumos.org/man/1/dladm not found

and unlike NetworkManager it wasn't fucking awful

The Solaris IP stack was. . . such a POS.

idk why illumos doesn't have it but docs.oracle.com/cd/E23824_01/html/821-1458/gkipd.html

Title: The dladm Command - Oracle Solaris Administration: Network Interfaces and Network Virtualization

Title: illumos: manual page: dladm.8

8? what?

there is no section 8 in solaris

but yes this dladm quote "(8)" unquote does describe the command i'm talking about

smf was... not very performant

SMF was great

it was like systemd but it wasn't fucking awful

(notice a pattern here??)

lol

alright the XML was a bit terrible

but you could implement it without that

does anyone remember the contracts API? *that* was great

i got some vaporware design scratchpad.pkgbase.live/Tn-dtb8OT--CttN1Djn8lw# that still needs an API that we don't have

Title: jrc: Principles - HedgeDoc

it was like Linux cgroups but uh... well, i hate to repeat myself, but it wasn't fucking awful. it let you track a group of child processes as a single entity

unixwitch: yes, that kinda thing is what we need. something orthogonal to jails that can do process tracking

i wrote a complete implementation of systemd-user (years before systemd-user actually existed) using contracts on Solaris

it even supported cron-like scheduled jobs

meena: since we've apparently accepted nvlists as a native API, perhaps people wouldn't object to importing contracts too

i wouldn't know where to start on implementing it though

same

i have to say i also preferred solaris event ports over kqueue but that's probably a done deal

there's quite a few things kqueue can't do, and possibly never will, like monitor a whole directory for changed e

*changes

oh i found my contracts thing, look at this: github.com/unixwitch/jobserver/blob…8b6fb86393b/jobserverd/sched.c#L230 - we can just use sigsend() to signal all processes in the contract, how awesome is this?

Title: jobserver/jobserverd/sched.c at e7229a1d5a2f800cb7d81af91378b8b6fb86393b · unixwitch/jobserver · GitHub

extremely awesome

you could even register a contract with event ports to get process notifications as events github.com/unixwitch/jobserver/blob…8b6fb86393b/jobserverd/sched.c#L313

Title: jobserver/jobserverd/sched.c at e7229a1d5a2f800cb7d81af91378b8b6fb86393b · unixwitch/jobserver · GitHub

soooo much nicer than SIGCHLD

well the move is finally complete.. next step moving the offiste backup and getting snapshots to work.. fun times still to go

freebsd doesn't accept patches via git send-email, right? it's either a PR with an attachment or github

(i just started working on another project that uses git send-email and i really like it, no need for other applications or web browsers)

no, please only bugzilla for ports stuff

i was thinking of src/

but i guess its true for both

Github or Phabricator

:-(

maybe i'll switch to openbsd, i bet they don't make you use a web browser

then again you do have to use openbsd

you can probably use freshports.org/devel/arcanist for phabricator, but I haven't tested it

Title: FreshPorts -- devel/arcanist: Command line interface for Phabricator

you don't *have to* use a browser… for submission…

meena: fish has a web browser?

i know about arc but i am extremely dubious i can patches committed this way. maybe i'm wrong...

it seems more like a tool for reviewing, while committing is left to the user

"For those lucky few with a graphical computer, you can set your colors and view functions, variables, and history all from a web page."

holy moly never heard of fishshell.com before

good luck with cvs ;/

unixwitch: "[…] to send code for review, […]"

meena: i am running 6 applications under Wayland right now and those applications are Alacitty, Alacitty, Alacitty, Alacitty, Alacitty, and Firefox, and i'm trying to get rid of Firefox.

there is some rumblings in the tmux area around this "replacemnet" zellij.dev

Title: Zellij

meena: Unix is a teletypewriter for me :-)

unixwitch > 9front ;/

Title: Contributing to FreeBSD | FreeBSD Documentation Portal

and there is freshports.org/devel/pybugz for bugzilla, but I haven't tested that one eighter

Title: FreshPorts -- devel/pybugz: Python command-line interface to bugzilla

nimaje: right, you can send code for review, but what happens after that? i don't have a commit bit, i need someone else to commit it

so far, github seems better for this than bugzilla

nimaje: i don't have enough monitor for such complex layouts

ye, it's always better if whole world is sitting on the one service/site

one shepherd said

meena: btw, i configure fish by editing ~/.config/fish/. i hope this annoys those so-called "graphical computer" people :-D

unixwitch: well, then someone else has to commit it, but that is the case with git send-email too

yay, cgit.freebsd.org/ports/commit/?id=4…762a319a037686c921accbb831db779715c

Title: ports - FreeBSD ports tree

nimaje: well, yeah, but if there a dedicated channel for send-email submissions, you'd imagine people would be around to commit them, like people spend time committing github PRs

whereas (i guess) there's no one sitting on top of phabricator reviews to commit them once they're approved, since phab is mostly used by committers

... just to be clear i'm not complaining, i was just asking a question. github is fine.

apparently there is a github cli, i should investigate that: cli.github.com

Title: GitHub CLI | Take GitHub to the command line

pkg install gh

nimaje: whoah. what does that mean? what things can be subpackages?

I need to see examples

random question (but this is freebsd-related) can someone recommend a large (several GB) BitTorrent file which is legal to download?

i'm trying to reproduce a filesystem issue for a PR

oh i guess this works: wiki.freebsd.org/Torrents

Title: Torrents - FreeBSD Wiki

meena: well, subpackages is "just" doing one build and then creating multiple packages out of that, besides the boring, separate out docs and headers and stuff for storage limited systems, some programs have a build where they build several frontends, like for example transmission, that gets especially annoying if those frontends need some common data files, before subpackages in that case the

packages for the frontends eighter had their own copy of the common data and conflicted on each other as a result or the maintainer had to create a packages for the data files and have the frontend depend on that (which could be annoying and in case the data is a result of the build, it is probably build once per frontend before subpackages); with subpackages the maintainer declares which files

get in which packages and how they depend on each other

unixwitch: Yes, a `...-dvd1.iso` should suit--the `amd64` one being ~4.5G.

torv: it seems like these files are not very well seeded unfortunately... but let's see

unixwitch: Give it some time for DHT. Worked for me.

the old review reviews.freebsd.org/D16457 contains git as an example at the end, not sure if that conforms the commited implementation, but it will probably look similar to that in ports

Title: ⚙ D16457 Subpackages. (deal with it.)

i thought ubuntu-23.10.1-desktop-amd64.iso might go quicker but it's sitting at 5KB/s... does no one use bittorrent anymore except for porn and pirated movies?

yes, i think that's true

(i do have a bittorrent link for 'debbie does dallas', which is in the public domain, but this is not probably not appropriate for a PR)

unixwitch: maybe try some other PD movies, like night of the living dead or plan 9 from outer space

Or Steamboat Willie!

i set the DHT port in rtorrent correctly which seems to have made the freebsd iso a bit happier, it's going at 16MB/s now

everyone is on about the steamboat willie

something is broken with my bridge(4) and IPv6 on -current even though the same config works on 14.0. it seems like freebsd doesn't response to ICMP neighbour discovery

how to get loginclass of a user in root shell?

pw user show $USER | awk -F: '{print $5}'

is it best way?

lists.freebsd.org/archives/freebsd-current/2024-January/005230.html in case anyone has an idea about my bridge(4) problem...

Title: bridge(4) and IPv6 broken?

VVD, possibly write a relatively short C program that runs getpwent and breaks out ->pw_class?

Ellenor, for start script for a port (software in ports)

why does your program depend on the login class?

not my

Ellenor, bugs.freebsd.org/bugzilla/show_bug.cgi?id=275851

Title: 275851 – [PATCH] databases/postgresql*-server: starting doesn't set up the class

okay...