rtprio: on 15.0?

do you recommend a certain scientific units (SI) conversion program for command-line and/or GUI that might also do old/Imperial (UK, USA, etc., which aren't always the same) units?

I believe the GNU project had one, darwin? gnu.org/software/units

thanks

i don't know if it understands imperial v. metric

er

imperial vs. USA

yeah; this is great: thanks!

it does these three types at least. Latin/Roman/Italian miles might also be different but I don't know who uses those

is the freebsd NTP pool broken? I can't get DNS lookups for {0,1,2}.freebsd.ntp.org using a number of resolvers

jmw: i think you're looking for [123].freebsd.POOL.ntp.org?

ivy: oh, duh :)

it's a bit confusing that ntp.org's DNS server doesn't seem to return NXDOMAIN properly...

actually, the output to units was a little confusing: doesn't actually say what's what

i get it now (reread)

mason: Thanks for the clarification. I think that the default actions.d/*.conf files provided with fail2ban might benefit a lot of a table-based design.

dacav: It's a very sleek way to do it.

I'll experiment a bit if I find time: if I get to learn it I might even consider contributing it to the pkg of fail2ban. I don't know if I've got the time for it!

(time is the limit, always :()

ivy: I've tried to submit the bug request we spoke about (/etc/dma/* permissions) but for some reason my report doesn't appear on the bug tracker: are they moderated perhaps?

dacav: as far as i know, no. when you submitted the bug it should take taken you to the page for that bug?

Then there was a problem in the bug tracker

what happened when you clicked 'Submit Bug' on the form?

I tried to find my bug in the list of bugs submitted by myself, but that's empty

It was yesterday evening, but I didn't notice any error: I assumed it was successfully submitted, and that it wasn't appearing because of moderation.

Too bad. I'll have to write it over again. It wasn't a book anywwy, so I'll be fine :D

after you click 'submit bug' it should take you to the url for the bug you just submitted, e.g. bugs.freebsd.org/bugzilla/show_bug.cgi?id=254445, if that doesn't happen please record what actually happens which will help someone (not me) debug the problem

OK, thanks. I suppose I can nag the person who answered my email in the registration process.

apparently if you use tmux as console in vm-bhyve, it creates a utmp entry for every vm: le-fay.org/tmp/7d/VRKQb5Ha6K ... i wonder if this is deliberate, or a bug

You guys know what happened to those port launchd efforts to freebsd?

I looked at the wiki page and it seems like theres a fair amount of repos started, some funtionality ported and then abandoned

did something happen, this is the second time in a week I've seen the launchd port mentioned

people are missing systemd alternative?

otis: surely launchd is not the answer

ivy: i don't dispute it. (i don't miss anything like that anyway)

anyway, i've just finished my "lunchd".

xD

initware got further before abandonment

initware looks interesting, but its lgpl licensed so i dont think any bsd could actually integrate it

... why? it's not as if it'd be infectious

Ellenor: at least for freebsd, such a huge amount of efforting was put into removing GPL-licensed software that i very much doubt any more would be added

LGPL != GPL

two different licenses, with two different effects on projects integrating software under them

at least in theory

i am very aware of the fact the the LGPL is a different license from the GPL, but i do not believe that makes a difference, both licenses prohibit distributing binaries without the accompanying source code

maybe I have misunderstood something but I feel like I have read somewhere that all the BSDs only ships with BSD-license compatible software

a common use-case for freebsd is taking the source, making some changes, and releasing the result as a commercial product, the LGPL does not allow that (if you modify the LGPL sources)

which is one of the reasons bash is installed via package manager

the CDDL has a similar restriction, but that's why ZFS remains an optional part of the system separated from the rest of the codebase

that compromise would never be accepted for a core part of the system like init

seti_: bash would never be integrated into base no matter what license it was under

yeah i had a feeling there would be something more to it even though i dont know what haha

the only reason we have tcsh is that BSD historically shipped csh, and tbh, i'd be happy to remove tcsh now that it's no longer root's shell

ivy, poor code quality?

I get why it isnt default but if we pretend it was BSD-licensed, why wouldnt it be given as an option?

Ellenor: what, tcsh? no, or at least i've never looked at the source code. i just don't think it belongs in base.

no, bash.

unless im mistaken you get a few shells you can choose from at install that arent sh

oh, well, same reason. we already have a shell, which is /bin/sh. if people want different interactive shells, that's what ports is for

Since the discussion is open: I understand that many user-oriented softwares are being more dependent on systemd lately. I also lurked a little in devuan, and it looks like they're sort of behind because of this. How is status in freebsd?

dacav: there is no systemd in freebsd and there never will be since it is intricately tied to Linux APIs like cgroups and namespaces

Yes, I know that.

But FreeBSD ships - say - KDE

as a port

what are you asking the status of then?

For example, I've tried devuan, out of couriosity

my understanding is that even gnome is not strictly dependent on systemd, but it is dependent upon some userspace apis that both systemd and some other programmes are able to provide

And it works, ...but! But suspend doesn't work, because of some systemd compatibility layer that is buggy

i had the impression freebsd's kde ports maintainer(s) have a fairly good relationship with kde and it's unlikely they will drop freebsd support

but hard to say for sure

Yeah, now I'm not a big fan of GNOME: I prefer KDE, or perhaps even a bare-bone window manager. Yet I think it's sad that people don't keep up with portability. I know, I know: preaching to the choir... :)

Good that KDE is inclined to support FreeBSD then!

dacav: i don't think you can directly relate a problem with suspend on a Linux distribution to FreeBSD, as they will have differently methods of doing that...

IIRC there are people on KDE foundation board who are also FreeBSD committers I'm sure cross-platform support is important to both communities

suspect/resume is historically quite tricky on freebsd to begin with, but if it doesn't work in KDE specifically, that would be a bug you could report

s/suspect/suspend

ivy: no, I understand that they're two different systems. But I could notice a certain struggle to keep devuan working when systemd is not around. Desktop environments are the same, and even if FreeBSD might have different compat layers, maintaining them requires effort for sure!

(in fact i don't even know how you request a system suspend on freebsd, but i know there is a way somehow)

also, suspend was just an example

well, is it possible KDE will introduce a bug because the person who committed some code only tested on systemd? ... sure, but it's still a bug, it can be found and fixed

Basically I'm a little sad that compatibilty across unixes is not always cared for, and honestly a little afraid too.

that's basically always been the case though

it's a shame, but it's how it is

Yeah, you're also right. I think that FreeBSD has a larger user base than Devuan anyway. Perhaps more resources to cope with those kind of bugs

it's funny because it's much easier to write portable software nowadays than it used to be (remember BSD / SysV?) but as a result people bother even less, because they only use Linux

Yes, there's definitely some contraddiction here

Most of differences are evened out even from development perspective (I'm more familiar with such side of it). E.g. most GNU Autoconf checks are pointless these days

do you have size_t here? :P

I hope im not poking a hornets nest here but is there zero desire around a new init system? I get that no one wants systemd freebsd-edition but some kind of more modern and appropriate init system?

there is, indeed, basically zero desire in mainline for changes to init. the launchd idea train came and went, without launchd even getting off the train.

IDK, is there something broken with the current one?

I've played some time ago with writing init files for a little program, and I found it very reasonable

from the user perspective I can't even tell the difference of course

(nor I couldn't before systemd)

seti_: you need to understand that this is a political question rather than a technical one

there are other rc systems that run on freebsd, changing the default is a huge

I understand

seti_: really every time i hear "more modern" i hear "i'm unwilling to learn what's already available", "poorly documented", and "in my fav trendy language"

i don't want fancy, new, trendy, or fun on my server. i want rock solid, stable, BORING, debugable, simple, repeatable.

systemd is none of those. :P alternate inits are often just not much better. optimizing parallelism for server boot time shouldn't be a big deal, because servers rarely reboot

ubuntu boots in a handful of seconds on my machines, ubuntu server in about 20 minutes

so I just ran ubuntu when I needed to

fast boot is super important for booting virtual machines

Demosthenex like I said, no one wants systemd freebsd-edition but thats not the same as a new init system

seti_: changes must be justified as better than what came before. what would a "new" init do better or differently?

Demosthenex: +1

and if there's a new feature needed, why couldn't it be added to the existing? why a whole new one?

stupid question... can bhyve run 32 bit operating systems or is it just AMD64?

seti_: i wasn't trying to shut you down. i'm curious what features you thought were lacking

Demosthenex I think the cloud people care about the boot times because they tend to destroy/create a lot of VMs. hasn't been an issue for me since I just run Jails with such workload and I mostly tune rc.conf as needed. most of my jails boot in less than a second.

antranigv: that may be, but if you're specializing your workload like that, you can minimize what you load at startup. it shouldn't take long

Demosthenex indeed, our fat supermicro takes longer to do POST than to boot the actual OS. POST ~10mins, boot ~30s.

exactly

Demosthenex I havent had any issues with freebsds init system or think any features are missing, I was curious about why the current init system seems to be unchanged even though it seemed like many wanted to try and integrate launchd or something to its effect at some point

seti_, that's because BSD isn't a SysV flavor. . . there's no run levels. . . unless you want to count single user mode and multi-user (normal) mode.

seti_: that's a fair question. and i'd suggest that mature features are left alone so we can focus on new features.

ivy: yeah on 15.0

If I want to run some server program when a jail starts as a specific user, is it rc I should use or is there something else?

mccd: depends how fancy you want to get. rc is certainly the "right" way to do it. you can also do something hacky like using cron and @reboot

I don't think user stuff should be done with rc.

That should be done in the user's own shit. . . cron is prolly a good place.

mccd didnt say it was "user stuff" per se

but it isnt uncommon to want to start a daemon as not-root

Well, maybe I glommed on to "starts as a specific user."

Well it doesn't have to start as a specific user, it's just a small node program that I want to run

so I could use su -c I guess and run the software itself as a specific user

@reboot works in user crontabs--dont need root involved

daemon(8) is also a nice avenue

Use of the reboot command is against best practices in almost every scenario.

okay this is weird, I've done bhyve PCI passthru like million times, why is it not working with this GPU? it's a "VGA compatible controller: NVIDIA Corporation GA104 [GeForce RTX 3070 Lite Hash Rate] (rev a1)". anyone has a guide for GPU passthru?

i don't believe it's compatible with gpu passthru

rtprio what do you mean???

CrtxReavr: can mitigate most of the stupid with lockf(1) ;)

i recall readying you can't passthrough gpus in bhyve

*reading

i think there is a wiki page that is outdated saying it doesn't work, but i think it does actually work in some cases just experimental and not yet documented

lol

cool

(but what i said is also just based on something i read, so that could also be wrong)

Hi. Quick question about ipfw. I have a rule: "65300 unreach port tcp from 192.168.122.1 to 127.0.0.1 80,443" -> This should be rejecting a http request from 192.168.122.1. It is set by fail2ban, but it looks quite straightforward to me. The requests however end up (according to the counters) on 65400 allow ip from any to any. Any idea on what could be wrong?

I'm requesting with curl -v 192.168.122.240

So I'm using /usr/local/bin/npm as my command in my rc file, but I notice it doesn't seem to run as a daemon

do I need to use the daemon command?

yes

wait, I got it! The rule is wrong!

the dst address is not correct!~

zi, reboot dosen't execute the `<rc.d script> stop` commands - it just sends everything a term and hopes for the best.

But you knew that.

I've just worked with plenty of software & daemons that requires a process more arduous to shutdown cleanly, than sending some process a terminate signal.

yea, again, im not saying it was the "right" thing to do but it is "a" thing to do

mccd didn't mention a whole lot in the details department

so...garbage in, garbage out is a possibliity

I do remember with Solaris having to 'reboot' when its IP stack vapor-locked, 'cause a proper shutdown command would just hang, require breaking out to OBP, manually fsck'ing, ad nausium.

'Course, Solaris x86 had no OBP, so that was anohter set of annoyance.

@reboot also fails if you ever restart crond

cuz that's considered a reboot

I've always loved how on FreeBSD you can just modify /etc/ttys and 'kill -1 1' to make it take effect.

CrtxReavr: That's the one I usually remember. There's also "init q" specifically for ttys file.

"6 SIGINT Reboot the machine"

On a BSD machine. . . sighs.

That's just dirty.

Sure is.

'telinit 6' yo!

I have a stuck pointer on elantech touchpad. According to gnu/linux, where it works, there's dual interface i2c and ps2. I notice that when I set hw.psm.elantech_support="1", evtest /dev/input/event6 does not display output when the touchpad is touched. However with hw.psm.elantech_support="0" evtest display feedback, and it works as a normal mouse, but sadly there is no scrolling. What can I do

to get more feel for what the problem is?

 

hmm, wg setconf keeps throwing "Line unrecognized: `Address=192.168.4.2/32'"

that is a valid address hmm

oh wait... its valid for wg-quick, maybe not for wg setconf...

Demosthenex: having a proper service manager, instead of hoping the pid stored in some pid file is still correct would be an idea