hi all

Hi.

any emacs/spacemacs experts in here. getting an error running emacs I cant understand why Im getting. would be awesome if someone knows how I can debug this.

how's 14 running?

drobban: maybe try #emacs channel

polyex: hm freebsd.org/releases/14.0R/schedule so a third rc was needed, but seems like currently they expect no fourth, or do they only add the "ass needed" entry for rc3 and add more rcs when they are sure they are happening?

Title: FreeBSD 14.0 Release Process | The FreeBSD Project

as needed sometimes

drobban: don't ask to ask, just ask. and please ask in the appropriate place. your question doesn't seem to be freebsd specific in any way. there's also #spacemacs on libera.

Hey, I was asking on Python channel but nobody answered. I have a problem with python3 and kqueue. I use multiprocessing and selector (defaults to kqueue) and receive an error. I use python 3.9.17 and FreeBSD 13.1. I also reproduced it on fresh installation of FreeBSD 13.2. Here is an example code with error message: bpa.st/6KQA

Title: View paste 6KQA

what do you expect that code to do? seems like nonsence to me, as the selector doesn't have any file descriptors

it's an example

If I do the same without the process it works correctly, in this case - blocks

hm, "The queue is not inherited by a child created with fork(2)." seems like you have to create the kqueue in the process

I am creating kqueue in the child process, s/DefaultSelector/KqueueSelector/

no, you are creating it in the constructor of Test, calling .start() creates the process and only what is in .run() is run in the child process

the better form is to pass a callable to the target kwarg of Process, instead of subclassing Process

(same with threads, pass callables, don't subclass, if you can avoid it)

oh, thanks, I'll try it

yep, seems that it's working

or just in general, composition over inheritance

on Linux I can do useradd -m username, non-interactively. What's the way to do that on FreeBSD?

I'd have to compare the manuals for details, but FreeBSD has adduser(8) which takes a similar syntax.

I look at the manual but failed to find it. Lemme read it again.

Shortish comparison: Put the usernames in a userlist file and run: adduser -f userlist

That'll create the new accounts using the defaults in /etc/adduser.conf

Which includes creating the home directory by default.

pw useradd

Note that lines in the userlist file must contain ten ':' characters after the username. Add 'random' after the last ':' to have it set a random password.

The adduser(8) command generally fills in empty fields between the ':' characters with defaults.

Okay, yeah, I keep forgetting about 'pw useradd'. That'd be 'pw useradd -m username' for the case mmlj4 mentioned.

With defaults in /etc/pw.conf

hmm...

so to get what I want on 13.2 I have to do: pw useradd username -m; chsh -s /usr/local/bin/bash username

thanks, I'd never heard of the pw command

pw adduser username -m -s /usr/local/bin/bash

one command

usually put -n before the username

so.. pw adduser -n username -m -s /usr/local/bin/bash

pw(8) is relatively newish, adduser(8) dates back to 2.1. Which release first had pw(8)?

hm, the HISTORY in the man page doesn't say :(

I do know it's been part of FBSD for quite a long time, though.

man.freebsd.org says 2.1.7.1

Yeah. Definitely been around since the 90's at least.

Huh, I missed the intro to it.

I'd imagine there's people here that weren't even born when it was intro'd. :)

Probably!

I've been doing batch user adds since 1994....

:D

I believe the first FBSD version I used was 3.1. Good grief that makes me feel old now.

... I got started on BSD 4.3 and Sys V R 3

Eh, we're just babies. :P

you rang? Unix Edition 7, on an m68k. Swapping, not paging

Nice! Must've been fun.

1995 baby here. definitely too young or nonexistent to remember it being intro'd.

why does the freebsd ami in aws take so long to initialize?

echelon: Eh? Rather famously, they're supposed to be down to the ms-range, I thought.

Oh, that was for Lambda specifically.

So it's limited to FireCracker.

debdrup: i don't think you can run lambda on top of anything besides amazon linux

echelon: Not sure I understand; Colin has definitely been doing work to speed FreeBSD up; latest tweet I can find mentions 25ms: twitter.com/cperciva/status/1694867769177813269

Title: Colin Percival on X: "@mfreeman451 Latest stats I've seen is around 75 ms, but those are from Firecracker performance testing and I don't know for certain if that's an apples-to-apples comparison against FreeBSD's 20 ms." / X

twitter.com/cperciva/status/1715887733414212016 even his laptop boots quick (5 whole seconds!)

Title: Colin Percival on X: "The FreeBSD 14 kernel takes about 5 seconds to boot on my laptop. Of that, about 1.5 seconds is ZFS running benchmarks on checksum algorithms. This seems slightly suboptimal." / X

debdrup: the vendor recommendation for instance type was t4g.micro, that's what i selected

echelon: Right

That's not the same thing as Firecracker, is it?

well, running it on a laptop isn't the same as running it on xen

Sorry, I got no real idea about anything AWS/EC2.

np

Xen is a lot closer to a laptop than Firecracker is, if I understand right.

i didn't bring up firecracker

Right.

Anyway, the point I was trying to make is that the head of the tree is significantly faster than the -RELEASEs that are out, and that a non-zero amount of those improvements have gone in before 14-STABLE was branched, so it should be in 14.0-RELEASE when that comes out.

freebsd.org/releases/14.0R/schedule Should be early November.

ok, cool.. looking forward to upgrade to it

Title: FreeBSD 14.0 Release Process | The FreeBSD Project

nice

this is how long it takes me to ssh in and get a prompt paste.ee/r/ObHJ5

One of the big differences between the Firecracker thing and either virtualized hardware or physical hardware (which amounts to the same thing, with the exception of the actual block diagram differences), is the hardware initialization step.

And the hierarchical nature of device enumeration means there's not a whole lot of speed-up to be had, from what I understand.

If there's speed-up to be had, it may be in the indivdual device drivers, but that's one of the things Colin invented TSLOG for, as described here: papers.freebsd.org/2018/bsdcan/perc…l-profiling_the_freebsd_kernel_boot

Title: Profiling the FreeBSD kernel boot: From hammer_time to start_init :: FreeBSD Presentations and Papers

Colin's been at this for over 5 years at this point, he takes it quite seriously :)

I'll also add that _none_ of his work involves init(8). :P

There's probably quite a bit of speed-up to be had there, but Colin's starting from the bottom, which if you ask me is the right methodology.

Inventing a directed acyclic graph for rc(8), ie. written in Almquist shell, might be a big ask though?

Could tsort be coaxed into spitting one out?

echelon: is linux much faster? iirc ec2 provisioning is just not that fast in general

also as I understand the image needs to be fully copied on ebs before starting the instance, so you should compare to a similarly sized linux image, not, like, alpine or something

Hello71: i can understand provisioning might take a while, but there's no explanation for why ssh takes so long

after it's already online

echelon, I am entering this discussion late but ssh has at least two things that sometimes feels like it is taking a long time.

1) DNS lookup of the connecting client. 2) key verification.

Reverse DNS lookup for connecting clients is sometimes slow or times out. This is often most of the time consumed when I look at things.

And then with everyone moving to 4096 bit keys then working through the host key verification and then the client key verification can take a while depending upon CPU speed.

rwp: thanks, that was it ;)

Which? DNS? Or ssh key handling? Or both? :-)

i dunno why it's even logged by default, for better details in logs?

dns

People. They like to see names of things rather than a bunch of address numbers.

V_PauAmma_V: I'm not sure I understand the question.

why are there both /usr/local/etc/ipsec.* and /usr/local/etc/*swan* files/directories as a part of strongswan?

how do i decide which one i'm supposed to be using

echelon: strongswan changed their configuration interface

google for stroke vs vici

so i can use either

if you're starting fresh go with the newer vici style configuration

the old one looks compact, but it becomes a mess if you want more complex things

well, the endpoint that will be connecting to it will be using the old interface

oh, spoke too soon, had to install another package for swanctl

Would (could?) FreeBSD be able to tap in to any sensor info that would tell me why this server's fans are spinnig up/down?

I guess most of that unless there's something already in place (I'm thinking there isn't since I would have to set it up..) would all be in the BIOS?

skered: sysctl should tell you some fair bit of info, and if not, there might be some packages in ports to help do that… or something else entirely

debdrup, I just noticed it uses rcorder instead. Perhaps I misunderstood your own question instead?

V_PauAmma_V: yes. and rcorder can also produce a parallelised version, but there's again, how do we consume unsay in rc appropriately?

What's "unsay"?

V_PauAmma_V: that's when you say something and you realize saying that was a bad idea.

I can't make sense of it that way in meena's answer. Perhaps I'm missing something.