alepzi: yes it's great if i want to run hikari 3 weeks from now

well you're running current, so lets you can fetch I imagine

also, potato CPU

jbo: i do take donations

you 2 should really room up to save money and buy more iron

alepzi: are you russian?

no?

oh, ok

you?

just curious because in russian they called hardware 'iron'

that was discussed prior

never seen that in other language much

ppl used to call it big iron

compiling ports on the new machine for the first time.... there go 600W

thank god this is all hydroelectric power

maybe we should get together and buy a #freebsd build server

that sounds nice!

company servers are still stuck on 2x Xeon E5-2690. they didn't feel slow before. now they do

jbo: does hikari do tiling?

lw, yes

i really hate sway but i like tiling

oh good

lw, keep in mind that I haven't used it in 2 or 3 years. I tested it on a system with intel integrated graphics (where it worked well) before using it on my desktop where it didn't work because Nvidia driver seem to use a different rendering backend. hence I am using X.

with bspwm :)

if i wanted X i'd just use twm, or maybe fvwm

sadly no one updated them for wayland yet

lw, does this still apply? forums.freebsd.org/threads/nvidia-current-status.81980

Title: Solved - Nvidia - Current status | The FreeBSD Forums

idk i've never used an nvidia card with freebsd

but... nvidia provides native freebsd drivers .__.

i know nvidia in general abandoned EGLStreams though

yes... but i don't own any nvidia graphics cards...

and anyway, what's the chance their drivers even work on 15.0

I never had a non-Nvidia GPU (except for intel integrated). But then again, I never had a non-Intel CPU.

why exactly are you running -current as your primary OS again?

why wouldn't i?

it has all the latest features, zero bugs, and i can instantly test my src changes

i would guess nearly all developers who submit patches to src use -current on desktop

but... daily driver? the system you earn $$$ with?

yes?

my home desktop & @work desktop are all running stable/14

i've been using it for 4 months now - zero crashes, zero bugs that impact anything important

main is really stable nowadays

before I switched to FreeBSD I was an Arch person... one day there was a bug where after exiting hibernate it overwrote / with /dev/zero .__.

(effectively, I assume it was more complex)

i mean you have to apply a bit of common sense, after some of the changes that went in this week i'm not going to update until next month probably

but updating 1-2 times a month is a pretty decent upgrade cycle

i think freebsd should get rid of -stable/-release tbh and just ship -current

yeah, I get that. but simply not good enough for my $atWork system

jbo: have you even tried it? how do you know it's not good enough

with bectl you can easily roll back a bad update, too

true - but I fear bugs that trash a zpool :D

yeah, but then 14.0-RELEASE corrupted my zpool so i'm resigned to low code quality of zfs at this point

might switch to netbsd raidframe or something, idk

raid/fs with built-in checksums seems hard to find

yeah I remember an ML thread from a few weeks/months ago about somebody voicing concerns over ZFS quality nowdays

i think i should just pick up a netapp/emc/something off ebay and go with that, as open source developers don't care to create reliable filesystem

lw, I have another question for you...

ok, i need a fag tho, brb

I was using emby for years now, generally happy, but lately I spend a lot of time ripping CDs (because diizzy). I feel like emby is not really nice for music. you used something else just for music, right?

the death and failures of zfs are greatly exaggerated in the press

realdeimos: i don't care about the press, i'm talking about problems i've run into in real life

14.0-R data corruption -> i filed the bug that resulted in the FreeBSD EN for this bug

native encryption broken -> i posted to fs@ about this months before Phoronix discovered the problem

ofc it's notable that i posted about that months ago even today, openzfs has completely ignored the issue and still tells people to use native zfs encryption

even though it's simply broken and should never be used

the more you spread the word about it the closer we get to fixes

we NEED native zfs encryption

bring attention and that'll bring help

+1

i don't want to be a cheerleader for "fixing zfs" i want a damned filesystem that works

jbo: for music i use Navidrome

lw it's so close it's way easier to get native encryption fixed than reimplement zfs

jbo: what's what the sublime-music port is a client for

just think of howmuch work is already done

jbo: (but it also has a web interface)

lw, "is it any good"?

I'm pretty anal about music

jbo: yeah i'm really happy with it so far

it's had no problem reading my mixed flac/mp3 music collection (1TB of files), all the tags seem fine

ack

sublime needs >=3.10, right?

it does support multi-disc albums which a lot of music players don't

y

but, there's a bit caveat

the version in ports is like two years old

lw, I'd like something with... *cough* an android client *cough*

i've been meaning to update it, in the mean time i run it in an alpine linux vm in bhyve

where the android client should support offline-download

jbo: i don't use android but i'd be surprised if there isn't a good android client

jbo: it uses an API which is compatible with another music server called "Subsonic", if that helps

so how about you provide a port patch and I take care of it? :>

i think Subsonic is pretty popular and has android clients

ugh

jbo: did you miss where i unmaintained all my ports

wut?

yes. I was pretty much absent the past 8 weeks or so

i can't express how little interest i have in dealing with that

I get that

I'm not a ports committer for fun :(

all of ports is just fucked, sorry, i understand you do ports and i'm sure you do a good job (and you really helped me with my stuff) but...

i'm just going to maintain my ports in a local branch from now on

yeah but you mostly touched python crap, right?

not by choice, but yeah

I mean... I get the criticisim (and point).

as I told you, I generally try not to touch python@. you made me break that rule more than once :>

i might be convinced to submit an update to navidrome

make my music experience great again!

jbo: what's going on with this tho bugs.freebsd.org/bugzilla/show_bug.cgi?id=276996

Title: 276996 – [NEW PORT] databases/postgres_exporter: PostgreSQL metric exporter for Prometheus

oh...

I...

will....

look at it.

sorry dude...

my bad

100%

it's fine it's not your fault

it shouldn't be your responsibility to get crap committed

oh it is, 100% in this case.

this is why i hate ports though :-P

nonono, it's assigned to me. it's not anybody's fault but mine.

let me testport this in 3 seconds

lw, patch is not applying cleanly (anymore?) can you update that?

jbo: literally your fault le-fay.org/tmp/30d/Screenshot.from.…024-04-17.at.01_39_56.835860150.png

i moved my monitoring to arm64 and didn't realise this package was missing

I'm sorry. It's my fault - no question.

wait what

how does it not apply cleanly, it's a new port

oh, the parent Makefile

jup

jbo: can you take a 'git format-patch' here or do you need it on the bug?

lw, PR would be preferred TBH

ok

how the heck do i download a patch from bugzilla

just to annoy you a bit more: this one I have to process through mentor approval. because I didn't do anything the past 8 weeks I didn't complete my mentee phase :<

oh, you click 'View'

lw, fetch bz-attachments.freebsd.org/attachment.cgi?id=248746

yeah

ok, what do i do here le-fay.org/tmp/30d/OONP9F.txt

the patch doesn't apply - ok, expected. but if i edit databases/Makefile, there are no conflict markers

I assume you can just resolve the conflict and git am --continue?

do i seriously have to apply it manually? what r u doing git

jbo: that's the thing, there is no conflict

it hasn't touched databases/Makefile at all

time for git-apply I guess

let me ask #git

pwned

this is going to be easier to just apply by hand, isn't it

sometimes you just gotta cave in

i asked #bsdmips, let's see if anyone there knows how to use git (i certainly don't)

ok

sorted, i think

i looked through 'git log' for a commit id from Feb 12th

checked out that version of the tree

now git am works fine

then git merge origin/main, conflict in databases/Makefile, easy to resolve

jbo: le-fay.org/tmp/30d/OveI8G.txt

oh let me put it on the PR

done

hmm, my zfs fileserver just rebooted for no apparent reason while running poudriere

weirdly no kernel dump

curious, any reason to use poudriere over synth?

lw: sounds like hardware problems to me

or rather.. smells like? do you know if it's panic'd before, and left behind a dump?

freebsd/poudriere a37a981f15f this looks pretty useful

Title: Add MUTUALLY_EXCLUSIVE_BUILD_PACKAGES · freebsd/poudriere@a37a981 · GitHub

hey all... is it possible to run a FreeBSD 12.x jail these days? I got an ancient app I need to house but it needs to run on older FreeBSD. How feasible would this be?

we allegedly turned off inet

hmm, nope it doesn't seem to have worked

cat /etc/src.conf: WITHOUT_INET=.

hmm do we have to do it this way ungleich.ch/en-us/cms/blog/2019/01/09/die-ipv4-die

Title: ungleich blog - Die IPv4, die!

:)

oh good, that doesn't work anymore

/usr/src/sys/netinet6/in6_pcb.c:569:16: error: variable 'inp' set but not used [-Werror,-Wunused-but-set-variable]

imagine not having to maintain any of this!

Well, if it was proprietary software, you woudn't have the option of maintaining it.

Imagine having bugs, but not being able to fix it.

You don't even need to imagine having bugs, because all software does - so it's easier on the imagination :D

debdrup: if you go that way, imagine that 99% of people in the world run proprietary firmware, which also has bugs :)

pkubaj: I'd rather not, that's a terrifying thought.

lists.freebsd.org/archives/freebsd-hackers/2024-April/003159.html is a good read.

Title: Re: Question regarding crunchgen(1) binaries

that was an interesting read

debdrup: our goal is to get rid of this code entirely, but for now some __unused is fine

okay it'd be nice if tools pointed out which file caused an error/warning

does sendmail even support ipv6

so it does but it breaks by default on an ipv6-only kernel we guess?

we could deal with the configuration issue... or we could patch libc.

tykling what jail options do you have for your poudriere build jail? ;)

im getting denied: mount: proc: Operation not permitted even though i set what i think is right in /etc/jail.conf on host

<class 'ImportError'>: /usr/local/lib/qt5/libQt5WebEngineCore.so.5: Undefined symbol "_ZN6snappy11RawCompressEPKcmPcPm"

17:01:31 /usr/local/lib/qt5/libQt5WebEngineCore.so.5: Undefined symbol "_ZN6snappy11RawCompressEPKcmPcPm"

Puzfire: did you set allow.mount.procfs=1 in your jail configuration

i think i did, but will revisit and make sure jmnbtslsQE, thanks for input

found the magic jail.conf in this post : dan.langille.org/2024/01/19/configu…g-poudriere-in-a-jail-on-freebsd-14

Title: Configuration for running poudriere in a jail on FreeBSD 14 – Dan Langille's Other Diary

now it runs

good :)

Hi, any Freebsd/Python expert here?

just ask your question

there is also #freebsd-python

Hi, I'm trying to run a program as root, but it seems it can't find the correct path to packages. As a normal user I've set these vars (pastebin.com/xzhZ8YYx) and it works ok. Now I want to run it as root (called from a program that runs from apache) but it can't find the right packages.

what are the values of the paths (especially $PYTHONPATH) for your different users?

The user that needs it is www (the script is called from CGI that runs under apache24)

this is a user that cannot login

I'm sure the PYTHONPATH variable isn't set for that user

you can set the PYTHONPATH variable in your apache config

mfisher: I'm reading about mod_env

it's probably not ideal for a long term fix (or in prod) but I've set a login shell and given the www user a home directory so that I could run scripts with a bit more visibility to what is going on

mfisher: Solved with mod_env

yay

I wonder if someone could help me, I have a application that displays a window, I have a bhyve linux instance and I'd like to the app to display on that instance but I keep getting: xterm: Xt error: Can't open display: 192.168.1.166:0.0

I've done and xhost + on 192.168.1.166 and I can ping it from the machine on which I want to run the xterm

crb__, I am not knowledgeable about X in bhyve so won't be able to help there but I can tell you why what you are trying now does not work.

rwp: ok, the bhyve client seems to be working fine and display local X programs

X is either not listening on the tcp port or there is a firewall blocking the tcp path or there is no tcp route for the path.

Oh! You have it working for you already. Great! Always glad to hear. :-)

rwp: what port does X listen on? there is no firewall unless centos has one built in and turned on by default, theses machines are both on a local network

crb__, Normally X will listen on the Unix domain socket at /tmp/.X11-unix/X0 which is protected by standard Unix file system permissions. That's the "new" way since around year 2000 or so.

Traditionally X would listen on port 6000+$DISPLAY (I think, or something similar). This was globally available over the network. The security dependent upon IP address of the remote host. Controlled using xhost.

is there a command I can use to see if anything is listen on that or other ports?

That was okay on a trusted LAN but as you might imagine it was also a great bit of fun on April Fool's Day when we would "meltdown" people, or send them crickets, or maybe Xsnow them.

But in recent years the entire tcp connection has been frowned upon strongly. X no longer listens to tcp by default (I think, pretty sure, not positive) and if one wants it to listen one needs to add -listen tcp (or correct option I would need to look up in the documentation) but don't do that because it's not a secure thing to do.

rwp yep I remember those days but today I'm on a local network that's behind a fire wall, and while I'd eventually like to lock this down more for the moment I'd just like to get any X connection going to this linux instance

Mostly these days we "ssh -X remotehost" and then let the X forwarding just do its thing.

Does that work for you?

It should be plenty fast enough on the local system.

ok so on the linux box I should run ssh -X remotest and when I get a shell on my remote host type xterm? What's the display value, is it just :0

crb__: it's assigned by ssh

So earlier you said "a bhyve linux" which implies to me that you are on FreeBSD with a bhyve VM and the bhyve VM is running Linux? So then you would need to ssh -X from FreeBSD to the Linux VM, right? Please clarify?

If your X server you want to display upon is actually a Linux X server then on Linux you can "ss -na | grep ^tcp.*LISTEN" to get a list of ports open for listening. And note that :6000 is not being listened upon (by default).

Warning: untrusted X11 forwarding setup failed: xauth key data not generated

Warning: untrusted X11 forwarding setup failed: xauth key data not generated

So... I hesitate to say how to generate that data (xauth generate $DISPLAY) because this seems like an XY problem.

One should never need to do that manually. It should all be handled automatically.

oops, too late I did that and now I can connect from the FreeBSD box to the linux box, but xterm -display :0 on the freebsd box still l results in xterm: Xt error: Can't open display: :0

I am still confused about which direction is the source and which is the destination.

top still have Warning: untrusted X11 forwarding setup failed: xauth key data not generated

rwp: sure, sorry. I want to run xterm -display on my freebsd machine and have the terminal window show up on the bhyve linux instance

And you can connect from freebsd to linux 192.168.1.166:6000 okay nothing blocking that tcp path?

And on Linux you can "ss -na | grep :6000" and see that something is listening on that port?

rwp nothing listening on that port on the linux side

And on Linux you can "ps -ef | grep X" and see something similar to "/usr/lib/xorg/Xorg -nolisten tcp vt07" but without the -nolisten option?

I don't know if I can connect to 6000 from the freebsd side but I can ssh that way (which obviously uses a different port) so I assume there is nothing blocking

If nothing is listening on that port then you will not be able to throw the display there over the X protocol. Because X is not listening there.

but I suppose if nothing is listening on the linux side it won't matter

On the Linux side if you really want to open X to the old traditional X display protocol then you would need to edit /etc/X11/xinit/xserverrc and remove the -nolisten tcp option. Probably changing it to the -listen tcp option. Then restart X.

See the "man Xserver" page and look for -listen for details.

rwp: thank you

FTR but I recommend not doing this and using "ssh -X" forwarding instead. That should Just Work and avoids all of the problems.

rwp: I don't see a file like that, I'll have to poke aronud

Obviously this is not a Linux support channel but I am curious what Linux distro you are using?

no it's not an I appreciate that, it's centos 7.9.2009

crb__: That CentOS is out of support.

Even CentOS 8 support is ending nowish.

Might be time to find a replacement.

I am sure that X still works the same as it did then though. And if someone is okay opening the X protocol port up then they are outside of the security circle of worry.

Fair point.

mason, it probably is but I use this instance for one VERY specific things and right now I'm just trying to get that app to run under linux emulation

perhaps I

perhaps It'd be easier to setup a vnc server on my freebsd box and export somewhere

If everything is local then "ssh -X" works very well. Even across a LAN. I would try it before discounting it.

It's across the Internet WAN where latency really kills X performance. That's X regardless of whether it is native X protocol or ssh -X or X via VNC.

I did try that Warning: untrusted X11 forwarding setup failed: xauth key data not generated

the ssh connects and give me a shell but also the forwarding failed

Hmm... I am not sure why it would do this. I am doing a little bit of research on it...

Hmm... I don't know. I'll just suggest a little poking around trying "xauth info" and then "xauth list" and hope there is an error with one of those that would point to something.

crb__ doesn't ssh have a configuration setting x-forwarding or something set to no?

s2r: I can't see that on either side in /etc/ssh/ssh_config

I don't know what the default is

"X11Forwarding no" defaults off on freebsd

s2r's suggestion is good. Double check both client and server side that X forwarding is not disabled. It's X11Forwarding on the server side.

i thought you needed xauth to do this

rwp: yes I checked neither side has a setting so it's whatever the default value is, and I don't know that

when you guys make and build apps for rollout how do you handle assets that are different depending on the env? like maybe you have a small fake user.db for the dev env, but for prod env it's a different file. do you store the files next to eachother? when you build for dev env, does it also include the prod user.db next to it still? would you ever wanna run an app with the assets of 1 env in the other env? like run production

build in dev env

crb__, So I have never needed this but reading the man page I am reminded that there is an "ssh -Y" option to "Enables trusted X11 forwarding. Trusted X11 forwardings are not subjected to the X11 SECURITY extension controls." which honestly I don't understand exactly what that means but I would try -Y at least once.

alepzi: not really a freebsd question

ya but lots of admins deal with the same stuff no?

I can only say that when building for production then everything must be as it will be for production. When building for staging for production everything must be as it will be in production. When building for testing for production everything must be as it will be in production.

ya good reality check. that ultimately ideally everything is the same no matter the env. so try to keep it that as much as i can

rwp: from the Freebsd box to the linux display sever?

ok so this time I didn't get an error about X11 forwarding

ok I got a local vnc server going and that seems to work

alepzi: on pattern i like is the ... config.file and you have config.dev.file and config.prod.file and it's up to you about copying/editing the right one

rtprio: ya i was thinking about that. so when you make a dev build, would it choose config.dev.file to import into the package as config.file, or would you package both up and the app code wouldn't choose 1 until runtime based on an env var?

there are dozens of ways to do it, but generally one instance only runs either prod or dev

so no sense in switching betwen the two

crb__, Since you are doing the thing you need to know the direction things are going. *I* can't keep them straight on your system! :-) But ssh from the X display system to the remote one.

when you name files for the environment they're made to run in, do you put it like user-dev.db, user.db.dev, user.dev.db or??

hi, there is a command line tool (and library) which is not NETCAT which is an old command line tool to implement script-based tools using networks/sockets/tcp/udp etc, but I forgot the name and netcat is so ubiquious that I can not find it. Anyone?

nc

i use ucat i think

rip: got it, thank you

rwp: thank you

crb__, Good luck!

uskerine, socat is another very similar tool to netcat, but more logically written.

It was not netcat and socat, but unfortunately I can not find it again

netcat or socat

You say script based and so of course perl/python/ruby/lua will do networking in a script. But I am sure that's not what you are looking for either.

I do think it was common and famous in the old days of UNIXes

No, it was an actual C written library+ command line tools

I'll mention "expect" just in case that triggers a memory.

it was actually pretty good, robust and tested, but no longer that common

It was not expect either

If I am working with SMTP then I will give a shoutout to the wonderful "swaks" tool, swiss army knife for smtp, as it is truly awesome.

this was more for regular TCP/UDP sockets, but intended to build services in shell scripts

I am not one of these people but some people have been known to preach the wonder of doing network programming in bash itself. I do not recommend it.

alepzi, This problem sounds like something very specific to your system. Meaning you have a lot of freedom to name it anything you wish.

On our OS the /etc/aliases.db file is always simply the /etc/aliases.db file. And same for the /etc/pwd.db file. They have already been named.

If you are creating your own files then it is a good idea to "namespace" them under your larger project name and not use too general or too short of a name. Ambiguity to be avoided.

For example user.db sounds too general for my liking. But postfix-user.db is something I would immediately know by looking at it what it is intending. (postfix is a mail transfer agent.)

General names and short names in a project named subdirectory is again okay. I would think /etc/user.db is something of freebsd but postfix/user.db I would know is something associated with postfix.

I am prototyping something and it is convenient, the library/command line toolkit was extremely good, I was surprised when I found it

uskerine, Sorry I could not help more. If you remember it please throw the name of it my direction because I would be curious. Until then good luck! I am AFK for a bit now.

thanks

might be this one

uskerine: what does prototyping something look like in that?

tcpserver and tcpclient are easy-to-use command-line tools for building TCP client-server applications.

tcpserver waits for incoming connections and, for each connection, runs a program of your choice. Your program receives environment variables showing the local and remote host names, IP addresses, and port numbers.

gosh, inetd

it is vintage certainly

uskerine: sounds inefficient but VERY handy for prototyping like you said

then you basically just use your same app logic in the final version

alepzi> do you really see like that? Maybe I am missing something but when I read this it looked to me like easier to nc

I am prototyping an internal LAN service for a few users

the invoked programs are normally either simple C programs or TCL scripts

the tcpclient tcpserver thing you described sounds really nice and easy

but that thing is so obscure now, I found it by mere chance

sounds like a ruby on rails app where you use a framework that lets you just handle requests and it handles all the connection shit

anything wrong with allowing the sftp subsystem on servers?

no less secure than ssh right? and handy for moving files around manually

i use rsync for automated files shuttling

Title: FreshPorts -- sysutils/ucspi-tcp: Command-line tools for building TCP client-server applications

nice you found it!

I do not need security for this, just doing simple requests that are served via command line tools

no, i was asking for myself

hehe

well it is clear why it is obscure, it is a bit convoluted to handle the redirection, but I still think this is a great to automate custom stuff on the shell (provided that it is a simple environment)

which way is right to enable bash-completion in my freebsd account? bsd.to/peBg

Title: dpaste/peBg (Plain Text)

alepzi, It's a bash specific syntax and for bash "." and "source" are the same thing. So it does not matter. Both are equivalent.

The ". file" syntax is standard /bin/sh and so bash gets that from there. The "source file" syntax originates with csh and bash pulls it in from csh.

rwp, alepzi left before you answered.