if i want to use latest repo, in pkg's FreeBSD.conf i change url to have latest instead of quarterly. do i do the same for the kmods repo url? kmods_latest_...

doesn't seem like that works

kerneldove, It is documented in the "man pkg.conf" file. It's the first two examples at the bottom of the man page.

"it" isn't documented, if you mean what the valid values are

valid values for the base repo are quarterly (default) and latest. but it seems like only quarterly kmods is valid?

it's valid to have latest for base repo but quarterly for kmods repo?

The example starts with FreeBSD-kmods: { url: "pkg+https://pkg.freebsd.org/${ABI}/kmods_quarterly_${VERSION_MINOR}",

ya

and?

And also url: "pkg+https://pkg.freebsd.org/${ABI}/latest",

did you even read what i wrote or just ignore it?

ivy there's a documentation opportunity: pkg.conf document valid values for the 2 repos which for base repo is latest and quarterly, and for kmods repo only quarterly is valid

fwiw *shrug*

I only provide the information that I have. I don't do miracles. Miracle Max is on break.

anyone have freebsd-update fetch fail pretty often on servers in europe? never ran into this before but till now i only ran from US servers

on dns requests it errors saying address family for hostname not supported

kerneldove: this isn't a fact, though

pkg.freebsd.org/FreeBSD:14:amd64 -> you'll note that there are in-fact latest and quarterly versions of the kmod repos

main only has latest

ah hm maybe the connectivity issue i'm having is unrelated

if i used latest for base repo and quarterly for kmods repo does that bork my install kevans?

1

ya ofc, piece of shit webhost, their network config was fucked up and that made it look like "latest" kmod wasn't working/valid

still think the valid values should be documented in pkg.conf *shrug*

is there a timeline to remove bios support from freebsd and only support uefi?

Hi, I was searching for some packages such as discord and found cgit.freebsd.org/ports/tree/net-im/linux-discord, does linux-discord or any app that has linux-app work through the linux binary ccompatibility layer on freebsd?

sozuba: yes, the discord client leverages linux "stuff" to get it to work

understood. Thank you :)

voy4g3r2 :

Yeah. AFAIK packages like that expect the Linux compatibility layer to be enabled. sysrc linux_enable="YES" vs sysrc linux_enable="NO" and you can test yourself.

i haven't installed freebsd yet, so can't check that yet. I am looking at all the packages i regularly use.

thank you o0x1eef

Ah, I see :)

Hope to install freebsd as a daily driver.

Good idea :D

tried it long back, had heating issues, hoping things work this time

:)

What kind of hardware do you have?

In the worst case scenario you could just use things like Discord in the browser too. In my experience, BSD is usually about compromise.

I use a Lenovo E450 laptop

Intel(R) Core(TM) i5-4300U (4) @ 2.90 GHz

16 GB Ram and 7400 RPM hard disk

o0x1eef:

Ah okay, Lenovo are usually good. I ran FreeBSD on an ideapad from Lenovo but wouldn't recommend that particular model. It didn't even work properly on Linux though. Usually thinkpads are also a good fit.

I tried freebsd more than a year or two ago. The laptop emmitted a lot of heat compared to linux and so i kinda quit and moved back to linux, i wanted to try and see if things are better now.

yeah i did read that hardware, even on well known brands are sometimes a hit and miss for certain stuffs when it comes to freebsd,

Indeed. Well, I encourage you to try and if you run into issues you can always ask for help here.

yup, will do

thanks you :)

yw

sozuba: vermaden.wordpress.com/2018/11/28/t…r-to-serve-freebsd-power-management

power management on freebsd is a dance and this is a real good source of information.. when you get furhter along in your journey.. he does a lot of cool things and shares.

Thank you. I think i tried this one a while ago, but can't be sure. Will try agin this time if i face similar issues. Note: This is the bug i filed long back bugs.freebsd.org/bugzilla/show_bug.cgi?id=271938 and the related forum post - forums.freebsd.org/threads/excessiv…d-e450-even-with-cpu-100-idle.89358

voy4g3r2 :^

Any thoughts on why some system calls are suddenly faster on FreeBSD than Linux? I have some measurements here: forums.freebsd.org/threads/since-wh…ve-faster-syscalls-than-linux.98957

I must have missed something, this didn't use to be the case.

were Linux's formerly faster? or just about the same?

most of those aren't syscalls

I referring to those calls that are syscalls.

even e.g. mutexes aren't syscalls every time, only if they have to block

I see getrusage at least. try disabling all mitigations on Linux. It could be that FreeBSD is missing some

a lot of mitigations for these recent CPU vulnerabilities involve doing something when crossing between kernel and userspace

Good point. Will do.

also, if you want to measure just the time to go to kernelspace and back, call getppid(). It's possibly the simplest syscall, it just gets an integer from a struct in the kernel and returns it. On Linux the libc is going to cache its result so only the first call is a syscall, you need your own syscall wrapper. On FreeBSD I have no idea

I find read() taking 75.9nsec total very strange, that's longer than it takes to go to the kernelspace and back on any CPU I've tested, the highest end such machine being the Ryzen 5 5600g

also, you're calling gettimeofday() to measure elapsed time. That's not the monotonic clock, your results can be influenced by NTP. You should call clock_gettime(CLOCK_MONOTONIC, ...) instead

Mitigations off in Linux speed up read(2) quite a bit. Numbers added to thread.

Thanks for the tip.

yeah, and the issue with that is that FreeBSD has its CPU vulnerability mitigations too, if you disable them it should also speed up.

we certainly have the same functionality, which currently includes gettimeofday and clock_gettime

I wonder why gettimeofday is no longer fast in Linux-6.17-rc2.

it turned into a syscall for some reason, 0 sys time on 6.8, and >µs on 6.17

and about my read() comment: I screwed up, I mean that's SHORTER than I've ever seen

seems freebsd doesn't have something like `mitigations=off` to disable everything for testing

you should change setitime(ITIMER_REAL, ...) to setitimer(ITIMER_PROF, ...) too

I pushed the clock_gettime(CLOCK_MONOTONIC, ...) change. I am not so sure about ITIMER_PROF. Don't I have to restart system calls in that case?

ITIMER_REAL should interrupt syscalls too. With enough iterations, the last syscall in the loop getting interrupted doesn't make a difference in the results

Yes. I guess I should polish up that loop.

hehe, Cracauer is here. Good to see you, man.

*was here

flatdog: still holding out.

good to see/know

Today's WTF: when starting a new tab in kitty (ctrl-shift-t), everything seemed to work normally except zsh didn't load any of its config. su to root and run zsh there, everything works fine. My own user? Nope. Run sh and then zsh? Still no config. The problem turned out to be that kitty had been deinstalled, and this weirdness was the symptom.

is there a timeline to remove bios support from freebsd and only support uefi?

kerneldove: i am not aware of any plans to do that

ok ty ivy. and what did you think of my doc suggestion for pkg.conf? to document the valid values for the 2 repos, that both can be quarterly or latest, and whether or not they can differ. (1 be quarterly, other latest)

kerneldove: i don't have any opinion on that, i suggest opening an issue in the pkg project

vkarlsen: is it a login shell?

oh. what?

anyone know how i can insert a "(" into a vnc over ipmi thing to a server?

sometimes the virtual consoles have a virtual keyboard you can use. sometmies it's really picky about timing--like you have to mash shift and 9 very quick together

ofc virtual kb doesn't have paren on it

Does it support copy/pasting?

some let you make custom macros, which could just be a (

that stuff rarely supports clipboard shit

ooi, what prevents you from just typing the character?

tried custom macro idea heh

im betting he's just getting a 9, right?