had a sandwich, managed to fire up (very manually)

how would I actually get the lua bootloader to pick up on a boot environment? surely it doesn't matter if they're called pool/ROOT/ or anything else, yeah{

?

the bootloader tried nothing and it's all out of ideas.

do I just forevermore have to put currdev in my loader.conf?

hi,everyone.This is my question. bsd.to/z6P7/raw

Title: z6P7

fengari, Seems like some wrong field or value type in "wpa_supplicant.conf" file

thank you,parv,i will check the file.

fengari, ... or "opp.conf" in your case

parv,I use wpa_passphrase to generate it,and the file is here. bsd.to/m93G

Title: dpaste/m93G (Plain Text)

fengari, Run "wpa_supplicant" with "-d -d" ("Enable debugging messages") option to see if that would identify the issue in English

parv, bsd.to/uEvv

Title: dpaste/uEvv (Plain Text)

fengari, Sorry I cannot decipher the "ioctl" error; debugging messages did not seem to help. To confirm, was the PSK generated with SSID (wpa_passphrase <ssid> <passhrase>)?

yes,surely.

sometimes i can connect correctly,sometimes not, idk why.

I am out of ideas

thank you, parv

That message from "ioctl" seems to be of no consequence: bugs.freebsd.org/bugzilla/show_bug.cgi?id=249542

Title: 249542 – 12.2-BETA wpa_supplicant makes ioctl[SIOCS80211, op=20, val=0, arg_len=7]: Invalid argument

I get those messages also and it seems fine

do you waiting for dhclient to get the link,when use dhcp, Reinhilde ?

that doesn't make any sense?

if you already have a link dhclient or rtsol should just work

sorry, Reinhilde ,i mean this, bsd.to/JnWF

Title: dpaste/JnWF (Plain Text)

it means your WiFi network is out of range or not letting you authenticate

check that the PSK in your wpa_supplicant.conf is correct, and (this is a really dumb one, but it stumbled me when I was messing with init systems) check that your WiFi interface is actually marked online

(ifconfig wlan0 up)

["you were messing with init systems? that doesn't sound like freebsd." and you're right - it's kinda not. However, I was messing with init systems in the context of the kernel and most of base from a fork of FreeBSD that by its nature uses the same rc system as FreeBSD.]

is it sufficient for loader.efi alone to be on the ESP, or must there also be other ancillaries to direct the loader as to where to find kernel etc?

thats pretty much it

how does the loader variable zfs_be_active get set? on my desktop, it seems like it doesn't (and I have to manually set currdev in ESP:\boot\loader.conf), whereas on my laptop, it gets set just fine and I can toggle between heterogenic (FBSD/HBSD) bootenvs that loader.efi knows how to boot just fine. I know I am holding it wrong, I just want to know how.

on the desktop the pool is called ultra, and on the laptop the pool is called zroot.

did jail.conf change ?!

Does anyone know why my loader in 13.1 insist on a serial console?

this will probably sound a bit non-important, but... is anyone running NsCDE on FVWM and having an issue with Firefox not using the motif style window?

Firefox did have the Motif style window decorations last week or so, but now it does not, and I'm wondering if anyone else has experienced this? Perhaps an update caused it to not use the default style.

I think I'm cursed

hi.... I have a swap problem.... I have added two more swap volumes and now I'm getting applications failing (being killed) when swap reaches just 11% of its capacity, how can I increase the kernel parameter to control the maximum swap I can use?

spydermocha77: ldd should show whether it's linked against any motif libraries

Reinhilde: probably

beastie: wild. how much RAM do you have? does the system know it's got the swap?

8Gb

yes...

at about the same swap space consumed as the amount of ram, the system starts denying more memory... and I also get a kernel message, which I need to search for, in order to show you the message.

meena: meep?

This is the kernel message: swap blk zone exhausted, increase kern.maxswzone

beastie: wait, how much swap did you add?

36Gb

yeah, i don't think that makes sense

i need that.... actually.

i need that to not need to close chrome while I'm using the last version of Netbeans.

I think a sensible cap is at 1.5. you can't really work the system trying to use swap as RAM

when it starts downloading the maven central index It explodes.

I need to maintain the programs started, even if they get swapped....

why does Maven need 44 G of memory?

I run it perfectly.... believe me... it's a core duo and it doesn't show any degradation (only when switching from netbeans to chrome)

why can't you memory cap Maven?

but each program has a footprint of 4.7Gb at the time the system starts denying memory.... now slow down in the system load... just programs crashing.

meena: I cannot say.... it starts the task "downloading maven central index" and at 87% netbeans crashes.

and the kernel just shows the message above.

worse if I'm running chrome in parallel.

chrome doesn't do any "not used tab swapping out" so it trusts the swapper.... so why the swapper cannot handle that... my freeswap shows all the swap space as in use.

the three swap volumes are at 11% full.

so, why not run the download on the command line while nothing is running, and only start netbeans when the maven index is populated

meena: probably because I don't know how to do that.

anyway, check sysctl -d kern.maxswzone

I'd even like to just deconfigure the maven server from the list of maven servers and just see what i have in the local maven repository actually.

when I need some library normally I search for it in maven central myself and add the dependency to the pom.xml file.... so I don't actually access the maven connector in netbeans.

but the maven index update starts asynchronously while I'm working... so I don't know until it is late and I've lost some work.

at some times netbeans is using over 4Gb of virtual memory itself... and chrome the same thing.

when the thing fails I'm using about 9-10Gb of virtual memory, which is only one Gb more than the actual ram I have.

beastie: maybe mvn dependency:go-offline works for your usecase, but I think maven needs the repo index to do its work not just for some optional features

nimaje: that's true.... but the problem is not with maven, but with netbeans :)

netbeans has a list of running services to acces the maven local repository and also has configured (by default with greyed menu options to delete it) an acces to maven central.

and at some point it downloads the complete index to make searches faster when you are searching for an artifact, but I actually learn to search for maven artifacts before I used netbeans and I prefer to add the dependencies to the pom.xml myself.

so the conclussion is that I don't use it... but I cannot deconfigure it.... I don't mind if it takes 2h or 2 months to update the index... but to see how netbans vanishes in the middle of something is annoying.

I would hope that maven caches the current index and downloads just what changed (and that netbeans downloads the index via maven)

no.... maven downloads what you need to have full access to the artifact, and the index is part of it (it creates a local repository under ~/.m2 with whatever you have 'mvn install' and its dependencies)

downloading (even just once) the full maven central repository is real nuts (they claim to use 110Tb of storage space)

not the repo, the index

maybe developmentality.wordpress.com/2010…-hit-take-maven-offline-in-netbeans helps you, but not sure what commands are affected by that setting

Title: Quick hit: Take Maven offline in NetBeans | Developmentality

thanks nimaje

but that still doesn't solve my problem, just avoids the trigger... I'll need to use more swap, and I think raising the limit to two or even three times the amount of RAM should be possible.

for now I've seen that the kernel message asks me to raise the kern.maxswzone parameter, but to change it I need to reboot, so I'll tell you later if it worked... now it's time to reboot.

thanks all for the hints and the good intentions, I'll tell you about the results.

Hi, I'm having trouble with a Keychron M1 mouse. It seems to only register input when the mouse is dragged down, produces aaabbccffgg, etc - depending on how fast the downward motion is. When I connect the mouse I get this - bsd.to/BsQa

Title: dpaste/BsQa (Console/Bash Session)

hi, solved!

nimaje: anyway, I take your hint of downloading the index offline (e.g. every night) so netbeans doesn't have to do it.

beastie: did raising kern.maxswzone do anything?

 

oops sorry

yes yes... it did it!!!

i enlarged it to three times its value (I was getting out of memory at one third the amount of ram + swap space) and it worked... then I found the parameter in the kernel source, so I got why the check is done.

the message output from the kernel allowed me to find the place where the test is done.

the thing is that the parameter controls the amount of metadata related to manage the swap pages, and the area allocated was exhausted.

but now I have downladed successfully the maven index while being able to browse the web at the same time.

the way freebsd manages the swap (IMHO) has no equal :)

beastie: I had used Linux for seven years before getting onto Solaris at work, and i had nowhere to grasp at anything to explain Linux's memory management. I learnt Solaris' within a week

:)

linux started imitating solaris in its user interface, but once it got windows, it started to shift slowly to Windows(TM) and now we have systemd on it.

I had been using solaris for years before I touched my first linux, and I learned linux in one week too!!!

Solaris (and Mac OS) has had a service manager for longer than Linux (but probably not longer than Windows). and it's better designed, and implemented, and documented. and it didn't cause a rift in its community

anyway, there's a working group for getting something like that to FreeBSD: github.com/freebsd/meetings/tree/master/supervision

Title: meetings/supervision at master · freebsd/meetings · GitHub

i understand your point, but I don't share it... the service manager is made for users that don't like configuring their systems. Las time I had to deal with systemd I passed over one week trying to know why I installed inetd to service some simple services and systemd insisted in putting his own version instead of mine.

I prefer to stay with init...

if that is going to be the future of freebsd, then probably at some point I'll go back to unix v7

you may have misunderstood me: I don't think systemd is anything to imitate.

of course not....

but I'm still using (in freebsd) mwm as window manager... resembling my old days with hp-ux

it's something you may need to provide compatibility shims for, because Linux has eclipsed posix as the standard. but great number a service managers on a large number of systems are far more useful and consistent than systemd

please no systemd-like crap, if freebsd ends up implementing some nonsense like that im going back to netbsd or openbsd

so do I

remember when we made sh the default shell for root, and people came out saying > "A is somewhat like V, which is the default on Linux, which is why changing C to A is exactly like introducing systemd"

im just throwing my opinion about it in the mix. i dont personally "hate" systemd, i just find the init scripts far more easy to understand and manage

i have refused to use vim, for that same reason, now I use vi

i use vim

I don't hate the idea of service supervision, I don't like the specific implementation found on Linux.

the only time I've lost completely a system (in 35 years) was when debian migrated to systemd.

init scripts worked fine, i still dont understand why the need/push to change it to something else

debdrup: i don't hate it for unexperienced users that don't want to mesh with the system... but it's not for me.

The thing is, service supervision needs a fault management framework of some sort.

You have to be able to know what kind of errors daemons experience.

that always allows you to tel the user "it's a systemd issue, you cannot avoid your computer doing that, it will do, either if you like or if you don't"

That's why the implementation in Solaris works, because they implemented Solaris Fault Management.

the push to systemd is why i'm slowly moving back to freebsd..

the biggest issue with the push is the fact that some packages need/want to install some sort of service into the system which automatically makes them incompatible with any other service-management system

gzar: that's an implementation issue

i personally just dont want an unaudited 2.2 million pile of sloppy code to replace time tested, secure services. It's not a good strategy for keeping machines secure

it's entirely possible to integrate fault management into freebsd and have it work with rc.d and daemon(8) as we have in the base system today; someone just needs to want to make it, and have the time/money to do so

sure but it still leaves the dev with more work if they want to support multiple service management systems

there is no standard for fault management, though

solaris is the only one to have implementede one

we don't ship our appliance at work for systemd based OSes. you want to run it there, you're on your own and if anything doesnt work, no support whatsoever from us.

Seems easier just let each service's init scripts deal with fault management , since each sevice will have different failures

we're gonna have to do our own

if freebsd makes their own, then we'll have 3 systems that end up incompatible with eachother

again, there's no standard.

Title: xkcd: Standards

welp, i'm done with this conversation.

'no standards' doesnt change anything about the situation.

there already are 2 systems, init scripts and systemd

2?! there are tens of them

Hi, I just upgraded to 13.2 and I'm trying an usb external Wifi adapter that shows up as: ugen0.5: <Realtek 802.11ac NIC> at usbus0

now how can I use it?

bugs.freebsd.org/bugzilla/show_bug.cgi?id=270709 is an example of a trivial thing you can't just fix once with our init: you can't just add an overload, you have to edit the script, and when the port updates you'll have to resolve the conflict

Title: 270709 – audio/oss Kernel panic at every shutdown after installing OSS

yep, it should be fixed in oss instead :)

s/fix/workaround/ then, my stand still points

is it possible to create a single drive zfs pool, then use the drive in a different controller (like a disk shelf) and keep the data in tact?

Just upgraded to 13.2 - Seems my file system (/) is quite full. I did a 'pkg clean' and it got rid of 2GB, but I cant figure out how to get rid of the rest

ncdu can help

Title: dpaste/7mTM (Plain Text)

debdrup, thanks

well

usually freebsd-update keeps a lot of files

and one can remove them after upgrade

and of course do not forget to remove all snapshots

gustik, thanks I'll man freebsd-update to see how to remove them

Thanks for the NCDU suggestion. Looks like the disk space hogs were locally installed programs in /usr/local/

Hello. zfs list creates this output: bsd.to/6Kgi

Title: dpaste/6Kgi (Plain Text)

Are the RELEASE-p* snapshot that get taken when upgrading between releases?

I am quite hesitant to run a zfs destroy to find out though :-)

Hi, i'm using syncthing, and the packaged version in latest is 1.22.0 which has know bugs

what are my options to upgrade it in a more up to date version ?

well

I do not think that /var/db/freebsd-update/files are in the man page

but you can remove it by doing 'find /var/db/freebsd-update/files -type f -delete'

thorre: I think they are boot environments, you can manage them with bectl

lvc: thank you for teaching me something new today :-)

hm

thorre: what does bectl list show?

meena: bsd.to/AGo1

Title: dpaste/AGo1 (Plain Text)

thorre: so unless you need to revert to one of those stages, i would bectl destroy all but two

thorre: the delte between the referenced data doesn't appear to be that much

thorre: you can convert them to bookmarks if you wish to keep the ability to do incremental backups as well

This means they no longer take up space for the changed records, so it still frees up space.

Does the KVM contain CWD of a process?

I cant seem to find it under 'struct kinfo_proc' - All I get is vnodes, and addresses in the kernel

RoyalYork: are you installing 13.2-STABLE? 13.2-RELEASE should be available on april's 11th.

ElectricJozin: KVM?

meena, man.freebsd.org/cgi/man.cgi?query=k…+and+Ports&arch=default&format=html

Title: kvm(3)

ElectricJozin: i see kvm_getenvv(3) but that's post-hoc

meena, did you take a look at the link I sent? its man kvm(3)

ElectricJozin, as the kernel has become more complicated, extracting this kind of information from a kernel dump has become more difficult

conceptually, it should be possible to use the name cache, but I don't think the code has been written

jilles, so I should use sysctl?

If so, do you know how to use the sysctl util to print the details first

The manual mentions no way of printing kern.proc.<pid>.pathname

ElectricJozin: I have seen the link. but i don't know what you're trying to accomplish, and in which context

meena, trying to get the cwd of a process in C (without using libprocstat).

man.freebsd.org/kvm_getenvv(3) isn't from procstat, it's part of kvm(3) and should give you access to the process' CWD if it exists

Title: kvm_getenvv(3)

(why?) is procfs deprecated?

(says it is in manual, but)

does getcwd(3) exist?

jgh: yes, why? it only works for the executing process

OIC. Some other process.

Reinhilde: because it was full of security issues

cor.

is it unfixable?

i can't remember the exact details, but i'm pretty sure it'd end up being close to a complete reimplementation from the ground up

all for a system that wasn't actively being used, and was only kept for compatibility

sysctl(8) was already been favoured long before procfs (something that started with the big rewrite of sysctl, which was in 2.2) was found to have the security issues, and the few things sysctl couldn't do were implemented in procstat

if i ever crack into it, I'll let you know where I get

if someone had the desire to make it now, i think it'd make sense to make procfs something that lies on top of sysctl

it's a bit of a mystery to me what it'd actually add to freebsd, but until someone proposes it as a thing they want to work on and intend to finish, i don't think that needs to be answered

aye

and let's not forget about freebsd.org/security/advisories/FreeBSD-SA-02:23.stdio.asc freebsd.org/security/advisories/FreeBSD-SA-01:55.procfs.asc and freebsd.org/security/advisories/FreeBSD-SA-04:17.procfs.asc

and freebsd.org/security/advisories/FreeBSD-SA-02:09.fstatfs.asc

i think that's all of them?

oh my cripes

there's a reason OpenBSD axed it completely

and when you look at what Linux provides vs what Solaris or FreeBSD provide, it's worlds apart, and we still think it's bad

wonder why bugs.freebsd.org/bugzilla/show_bug.cgi?id=230620 is listed as "needs special attention"

Title: 230620 – "install -d" issue

very interesting bugs.freebsd.org/bugzilla/show_bug.cgi?id=270728

Title: 270728 – Clock stops, but system keeps running

meena: probably

meena: what is your current idea for how supervision ought to look?

OpenBSD also axed the kernel dynamic linker entirely.

anyway. I'll stop bothering you and keep hacking on my ifupdown-ng executors.

Reinhilde: Are you doing ifupdown for FreeBSD?

mason: Rudimentarily.

Nice. That'd be a win.

--- wait what

I've just been doing it as a shitpost in software form and to see if I could

And if you can, tidy it up and get it into ports. :P

Ports? It seems to like the idea of polluting /etc and /usr/libexec (and on my own system, I have been all too happy to let it)

ixmpp: did you read the stuff in github.com/freebsd/meetings/tree/master/supervision ? the current idea is: kqueue / kevent (because it's easier than signals), see freebsd/freebsd-src #701 and then explore (dynamic) login classes as a basic container

Title: meetings/supervision at master · freebsd/meetings · GitHub

I also haven't yet done the integ work to make an rc script.

Reinhilde: why not integrate it into netif?

it's probably just 8652 lines of shell

I swear to the gods, if FreeBSD ends up with systemd, i will personally stab everyone who was involved :( FreeBSD and I have been friends for decades now. At least since late 1990s

let me reiterate: "A is somewhat like V, which is the default on Linux, which is why changing C to A is exactly like introducing systemd"

i'm used to openrc and s6, on linux, neither of which are horrible

please, for the love of piss and all that is golden, go out into the world, and look at Service managers that aren't systemd. service managers that work without being the only fucking thing people talk about

:-)

imagine if all everyone ever talked about how SMF was bad, because it used XML for some parts. and not, say, dtrace, zfs, zones, boot environments and a million other engineering feats

macOS had launchd, and somehow that's not the only thing people talk about

Windows has an all powerful integrated service manager and somehow its not the only thing people talk about

meena: netif is a bit like jumping from a cliff without a 'chute

nobody talked about them on linux either until pottering shat all over it

Reinhilde: that's a fair assessment.

270728 – Clock stops, but system keeps running bugs.freebsd.org/bugzilla/show_bug.cgi?id=270728

meena: ¿

i asked because i'm interested in how it'll turn out, not because i fear systemd

if it's closer to openrc, or much thinner, for e.g

s6's idea of dependency resolution is a funny one

basically "fail to start unless all dependencies are running", then, launch all services at once so they eventually all resolve

but actually, i have trouble faulting that tactic

ixmpp: we don't know yet, but we think it'll be a combination of jails and login class API to approximate cgroups / Solaris contracts. and we want the core to be in C, the actual rc in Lua, so people can easily extend / debug it (GNU shepard can give you a REPL, and i think we could too) and configured in UCL, so, a FreeBSD native

Good.

neat

makes me want to fire up linux 2.6 and patch up daemontools.

anyway.

either way, that's the rough ideas and tools we have in mind now, but who knows where we'll end up, actually

if any where

parv: do comment

I am convinced by the evidence that the wheel is being reinvented.

?

Reinhilde: the wheel was "reinvented" about 7 times since its inception, and if it wasn't, you wouldn't be getting any where

anyway...

do any existing service managers even work, on freebsd anyway?

its not reinventing if it doesn't exist :p

the first revolution was spokes, and I'm pretty, and sure people where saying back them that the wheel was being reinvented, too.

ixmpp: openrc, but just barely

do i have someone on ignore who i shouldn't do o.o

ooh now i really want to see if i can build a system out of that

ixmpp: welcome back to the land of the living

you blocked me? why

taking this off channel

meena: UCL configuration pleases me greatly

imagine having things in… not global scope

must be comfy

there's a nice trick on linux, subreapers

helps keep processes where they should be

freebsd has that too

FreeBSD is also capable of that I believe, ixmpp,

can that be approximated on freebsd in any way

oh good

(how?)

trying to remember

procctl!

daemon(8) does it

gotcha

so yeah. The ifupdown I am basing my implementation on is ifupdown-ng, which is Alpinelinux's reimplementation. The initial prototype took almost no changes to the actual C code; some headers are missing and exactly one file has outright Linux-specific code (which module I disabled to not have to port). The real work was in creating the new executor scripts, which I made an utter hatchet job of,

only did the bare minimum, and am still not sure how to handle with respect to wifi.

it also looks much tidier than the linux way, wonderful

I can't remember if proccontrol(1) (which should be in section 8?) can do it

signs point to no

hm, then i can't remember how it's done, other than daemon(8)

I've upgraded a package and found some secondary failures, I'm wondering if there's a decent way to downgrade a package from `pkg`

rtyler: old package still on cache?

rtyler: you might have the old version in /var/cache/pkg/ that you can install using pkg-add(8)

you might need libmap.conf(5) thogh

though*

lemme check

also i recently found out that libmap.conf(5) is not a thing on a lot of other unix-likes, and i don't know how anyone else makes things work now

that's from the ELF loader, yeah?

meena: deamon(8) is not acting as a subreaper (yet)

that is on my todo list

drats, I found one package in /var/cache/pkg that I need, but one is missing

I was able to walk one package back, just need one more

unfortunately the mastodon package is not yet ready to accept the latest ruby31 package :(