there were plenty of articles on networking and such things back in the day

GUI, Networking, Security ... big three items

he mades several predictions that proven to be right later: that every single computer would end up having a network card which would cost 0$, that whatever UNIX evolution would exist in the future, the important thing would be the capability of running several processes in parallel, but that the multiuser would be irrelevant, and he made other comments that I do not think they were obvious back then

it is old stuff, it seems from other age. Many of the contributors are no longer there too (which is obvious). There was another article about the director of research of AT&T talking about networking, convergence of Voice and Data and things like that

mns I am reading 1985 now, so the GUI is still not that much a thing -yet-. There are comments about the BLIT and there was a lightweight window manager I do not recall the name

MGR

Title: ManaGeR - Wikipedia

uskerine: you're going down a rabbit hole of fascinating history :-) You can get a bilt emulator with 9front, a Plan 9 fork. en.wikipedia.org/wiki/Blit_(computer_terminal)

Title: Blit (computer terminal) - Wikipedia

Is FreeBSD also affected by this data loss bug? openzfs/zfs #15526

Title: some copied files are corrupted (chunks replaced by zeros) · Issue #15526 · openzfs/zfs · GitHub

15526 – [NEW PORT] security/pgpgpg: a wrapper for GnuPG to emulate PGP 2.6 bugs.freebsd.org/bugzilla/show_bug.cgi?id=15526

And is there a way to update to a version that isn't?

It sounds pretty serious

Especially because a scrub doesn't see it

wildeboskat: Look at the comment with: "With FreeBSD 14.0-RELEASE, there's complementary use of the vfs.zfs.bclone_enabled kernel state, 0 (zero):"

Hmm I don't really know what that means :)

But from what I understand from the hacker news thread, the bug also appears with block cloning deactivated, it's just much less likely

For example here: news.ycombinator.com/item?id=38405732

Title: The root cause of the bug may have been present for a long time. It increased in... | Hacker News

But his "quick workaround" obviously doesn't work on freebsd as it requires sysfs

On Linux they do this: echo 0 > /sys/module/zfs/parameters/zfs_dmu_offset_next_sync

But on FreeBSD there must be a corresponding sysctl I imagine

But not sure which

hrm, I thought the bug only appeared with block cloning, but I'm insufficiently familiar with it

vfs.zfs.dmu_offset_next_sync

Thanks!

This was on for me :(

On 13.2

The issue is neither related to BRT nor DMU offsetting, it's _probably_ related to SEEK_HOLE. All BRT or DMU offsetting seems to do is make the race-condition that's the source of the bug more likely to occur. For what it's worth though, outside of specific reproducers, I haven't actually seen anyone make mention of running into the issue on FreeBSD systems, and I'm not sure I understand why.

A change to the GNU coreutils package, which is used by a lot of Linux distributions, made it much more likely to happen on Linux, and it's not like it _can't_ happen on FreeBSD (see: the reproducers proving that), but it doesn't seem to without running them?

I can't imagine that there won't be an EN for it, but it has to be tested properly first and that takes time.

It did make me wonder whether KCSAN (the Kernel Concurrency SANitizer, ported from NetBSD to FreeBSD) would've caught this, because it seems like the exact sort of thing that that's made for.

There's a report of a build failing for a file containing zeroes: echo 0 > /sys/module/zfs/parameters/zfs_dmu_offset_next_sync

Oops

Title: The recent FreeBSD 14 release apparently failed to build for one of the platform... | Hacker News

That's the right link

But thanks for the explanation debdrup! Do you know when a version that's fixed is pushed? Or does it require more testing to determine the "probably" part?

wildeboskat: I can't imagine that it won't be part of an EN, so I'd recommend subscribing to the announcement mailing list on lists.freebsd.org/subscription/freebsd-announce

Title: FreeBSD Mailing lists: subscription for freebsd-announce

Ugh I really hate mailing lists clutting up my mailbox, but I will keep an eye on the web part of it, thanks!

announce@ is moderated, only people with explicit permission can use it.

Ah ok

I see

But this sounds like a big enough issue to warrant an EN?

What does and doesn't get an EN isn't really codified, but like I said, I can't imagine an EN not getting issued.

Ahhh sorry I understand now, I thought you meant to say that it wouldn't be part of an EN which surprised me

Thanks!

announce@ gets less than 10 mails / month (with a few exceptions where there were 11/month).

Thanks! That's pretty decent :3

wildeboskat: that's probably on me, so I apologize.

Nono I reread it and you said it correctly

It's 5am here so I'm not at my most observant

fwiw, KCSAN wouldn't have detected it. its was a logic error - one logical operation dirtied two different kinds of things, but lseek() was only checking one.

under what circumstances do I need to be worried about this ZFS bug ?

Well apparently block cloning and DMU offset (whatever that is :) ) make it more likely to happen

But don't take my word for it

I'm not really into filesystems

I jsut have defaults for ZFS

you almost certainly do not need to worry about it.

and if you set sysctl vfs.zfs.dmu_offset_next_sync=0 , you definitely do not need to worry about it.

Thank you

bugs.freebsd.org/bugzilla/show_bug.cgi?id=275308 is the EN tracking issue

Title: 275308 – EN tracking issue for potential ZFS data corruption

What I find the most worrying is that it's silent - the features which are supposed to detect corruption don't

veg: how did your netatalk stuff go?

yeah. fortunately its wildly difficult to hit on any non-contrived workload, and even harder to hit without the application noticing

debdrup: fwiw i'm pretty sure that exact bug is what broke the PINE64 build for 14.0

some folks had been hitting it intermittently in kernel builds

oh, mentioned later

kevans: You're referring to this right? news.ycombinator.com/item?id=38408993

Title: The recent FreeBSD 14 release apparently failed to build for one of the platform... | Hacker News

Oh yeah I just dug down into it and it is jndeed the PINE build

yeah

That sounds pretty real-world to me o_O

robn: oh, hey, how long have you been hiding out in here? :-)

PS not trying to detract from FreeBSD or ZFS, just worried about my data lol

100+TB is a lot :D

wildeboskat: I said "difficult to hit without the application noticing". in this case, the application noticed

Ahh good point

wildeboskat: I'd put money on you not having any stored corruption caused by this bug

Thanks :3

I might lose, but I'll almost certainly win :)

Most of my data is actually not on ZFS

Though I was planning to move it, but I'll wait a but until all this is clear

It is mostly on ext4 with the idea behind it that a simple FS is easier to extract data from if it screws upw

kevans: here, maybe a year. my history of lurking in irc is more like 25 years lol

However the lack of block-checksums etc is something worth going to ZFS for

wildeboskat: all other things being equal, I'd prefer to be recovering from ZFS than ext4 (or any FFS-style filesystem). even the worst failures still leave a lot of structure on the disk

Really? I didn't imagien that

but I recognise the tools and expertise aren't as readily available

ah, nice- always nice to see zfs folks further infiltrating the freebsd community, whether for zfs purposes or not

Yeah I wouldn't know how a ZFS works, ext4 is easier to understand

yeah, coz ZFS never overwrites anything, only copies. including its own housekeeping structures. so if you totally toast the current version of something, a previous one is likely to have something you can use

this is going to sound crazy: Have ZFS users, who use the /compressed option on a mount.. notice performance lose with processing files/

i got time machine to work on mac, through netatalk, but it is HELLA slow.. like in 20 minutes, i have copied 25 megs of data...

(also snapshots etc out of the box make it easy to make even _more_ copies)

well scratch taht question.. it is doing "burst" uploads.. it has to be the client and not the server.. never mind

kevans: this year I learned that I've always been a freebsd-er at heart, I just accidentally ended up doing linux for 20 years and being depressed the whole time :D

Linux had a few good years.

wildeboskat: I used to be on linux, using ext3/ext4 and got some corruption. freebsd was the only OS from which I was able to recover that data. I tried NetBSD and a bunch of other linuxes to no avail. After that, I just switched to ZFS, have had numerous power outages and such things, but no data loss. once you get the hang of some of the basic stuff, its easy.

Linux is _fine_. I just think we deserve more than that

yeah, all operating systems/filesystems/disks will eat your data sometimes. can't do much about that as long as physics exists and humans are fallible

zfs at least gives you a fighting chance of knowing before its too late

I'm a FreeBSD user by the way, it's just my fileserver is still on Linux because I couldn't find a great BSD replacement for mhddfs

oh for sure. my daily driver is still linux, because I've been doing it for decades and its very hard to find time to disentangle it

not that I have to. all friends here

(it is running zfs ofc)

anyway I should stop shilling for Big Storage and get back to this sodding bug

robn: aha, so what I'm hearing is that we have enough upside that we could conceivably annoy you with enough quality of life improvements to implement that you eventually accept a committer branding

oh quite easily. I'm already on that road; not hard to accelerate that by sniping me with interesting tidbits

... so long as zfs doesn't keep catching fire and taking up all my free time, anyway

eheh

What is FreeBSD like these days, thinking of switching again to F'BSD

like for desktop?

AumShivaya: it's pretty fine

cool la_mettrie

Does Steam work on it yet?

with Linux Compat or something?

i'm not sure. zork, lgeneral, freeorion and vms-empire are enough to me. maybe somebody else here knows the answer

what is a zork?

I am migrating from freebsd to debian but my disks are encrypted via geli that is not in linux. Should I use freebsd (as guest) inside virt manager and access those drives?

so i want to give the fbsd server some swap but am trying to think of the best way to do this. i guess i have these 2GB partitions.. is it common practice to mirror them?

Beladona: i would think, yes. You can't access them from non-BSD env

Beladona if you can't get answers here go to the lists, forums or discord. Everything should work not just one place. Of course you will need to wait for an answer

jb1277976: I love fbsd but cannot make manythings work. I have to make this choice

angry_vincent: ok

angry_vincent: any thoughts on migrating current freebsd in to that guest virt manager based?

Hi, upgraded to 14.0-RELEASE yesterday and everything is working fine now. During the upgrade I noticed that the root users shell no longer is /bin/csh in a passwd merge. I did not dare to change the root shell during the upgrade so on my machine root still has /bin/csh instead of the now default /bin/sh

Is it recommended to change to /bin/sh for root?

hi i have a anano pi r4s with realtek NICs that i ran freebsd 13 + opnsense and the iperf3 speeds were only 500mbit max? Have things changed with freeebsd 14 and realtek NICs?

s/anano/nano

Is there anything that freebsd do for zfs and not zfs FS itself? Ie something OS dependant. Eg checking disks automatically itself if corruption happens etc. I am migrating to debian but will use zfs. So that is why I was asking

thorre: use whichever shell you prefer, I really doubt csh will be removed from base any time soon

the reason for the change is most likely that 95% of *NIX/POSIX users expect Bourne shell, not C shell

Remilia: I have the same feeling to but I wanted to make sure ;-) Maybe someone knows something that I do not.

yeah, lots of crusty people (like me) prefer csh as their root shell

I cannot deal with csh haha

crusty people ... made me smile

but, 90% of my csh usage follows from jexex -l foo csh

I remember when working with sunos and the "cool" shell to use was korn shell, "ksh -o emacs" for esc-completion.

thorre: my first non-handmade PC was a SPARCstation with SunOS and I am 100% sure the default was the Bourne shell

it was the same on my second PC, an SGI Indy

cool people sometimes deviate from defaults

Yes, sh was the default. Ksh was included from Sun. And bash could be installed via cdrom ;-)

mine was the first SPARCstation I think

it is somewhere in the basement back at home

I used bash cuz i didn't know better, but then again, i only came to sun at the 9/10 era

together with PC-8801 and various PC-98xx

I remember when a new hire asked me how to turn the machine off after he worked for us for several months. There was an email about weekend maintanance and he had never shut down his work station before

I still use ksh -o emacs

paulf: I migrated to zsh when that was made the default in macOS. I could honestly not see me having a shell without fuzzyfind and other nice things that zsh has.

But it depends on your use-case.

in retrospect upgrading to 14 was a bad decision

thorre: you can keep csh of course, the point was to change the default

Beladona: sorry, i cannot help with migration. i never used debian and quite possible never will. anyway, not a topic for this channel.

Remilia: you have to suffer so that others don't

yeah, nobody is talking about deprecating csh

sams: Realtec NICs are genuinely terrible so you are likely getting bottlenecked on performance

pstef: I will change to /bin/sh since this is the new default and I barely never use the root-user account anyway. When sudo is not enough I usually use the toor-user with zsh. root for me is only for rescue scenarios and there it can be good to have the default.

angry_vincent: ok

Remilia: any chance you're monitoring can record processes are running during those high load periods and what their stacks look like?

can someone offer advice on how to track vnode exhaustion (possibly leak?) because I really do not like the kernel spending 4% of its CPU time in locks and another 100% in vnlru

put differently, can we can figure what starts it?

meena: I am not good enough to figure that out

angry_vincent: what desktop OS you use?

I know it is definitely something in periodic

and in periodic, you have find

which likely loves vnodes

oh, I could move periodic runs to a time when I am awake

I'm thinking of something simple, like a cron job that records what's happening at the time in a text file

or that lol

sams: I don't claim to understand networking, but just recently I had to disable LRO on the FreeBSD side to get speed from 6 MB/s to 90 MB/s (over a gigabit link)

meena: you know what was the worst after I tried increasing max vnodes, the way jails went slow motion

`service jail stop X` just stalling forever

now that sounds like something you can take lots of nice dtrace snapshots of

hotkernel, pstacks, etc

meena: yes but that is just the consequences of vnlru spending time in locks

looked at the load graph carefully and it is definitely periodic, spike starts at 03:01

:sicko: Oh yeah! 3a.

moved periodic to 13:00

now wondering if it will trigger today

meena: if I changed the crontab line that already ran today, will it run at the new, later, time?

I am assuming yes, doubt cron cares

not like it can tell that I edited a line and not removed and added

Beladona: i do not know what is desktop OS. to me this sounds like artifical word construct. i use wayland compositor.

angry_vincent: Desktop OS: an OS designed for good desktop support. E.g GUI, bluetooth, hardware, wacom pen (muy case of issues)

etc

In my experience, there's such a wide amount of variance in what people want and expect from a desktop, that just about nothing can be generalized.

debdrup: a few things are thought to be basic. Eg bluetooth

About all that can be said is that FreeBSD can be made to work well as a desktop OS, if you've got the time, patience, inclination, and technical knowhow/willingness to learn.

Beladona: I have bluetooth on my T480s working just fine.

Tested with both a mouse and a headset.

Many do. At what cost? How much time?

Beladona: that's hard to quantify; how long does it take for a power-user to configure _any_ OS?

An unenumerable amount, because powerusers tend to always tweak something or other.

Agreed but some OSes do it for you. Built in. Better hardware support out of the box

If it doesn't work for you, don't sunk-cost it - just accept that and move on.

so millions of power users don't have to spend any second on it

1 second X a million users = 1 m man seconds.

Where did you get the idea that the FreeBSD project or community wants to market itself to millions of power users?

when I reinstall Windows from scratch it takes me around 2 days to set it up just right

Few people want to just get the job done quickly. Despite being power users, they don't want to spend time.

so I dunno, I feel like for power users there would not be much difference

Well if not millions, thousands?

Beladona: it sounds like you're conflating what would make it easier for you as being equivalent to what would make it easier for everyone.

I am relating my feelings as a human and assuming that many would feel similar "getting the job done in less time and expecting many things from OS out of the box". Specially if the same is done in many other known OSes. Freebsd had been the best OS I ever used. Except the things said above.

So no bad feelings about Freebsd. Just exchanging thoughts

setting up a FreeBSD desktop system for me would definitely take less time than with Windows but maybe you mean Mac OS?

I last used that on a Powerbook G4 and remember nothing

Remilia: how many times have you set up Windows, for comparison?

If anyone is a power user, it does not means he and ALL others like him (thousands if not millions) have to spend time on it. VS that be made as in the core of OS by OS devs

debdrup: I would not be able to count, likely >200

actually scratch that

>500, counting from 3.11

When I was a Windows toucher, I would routinely install Windows often enough, that to this day I could probably still do it without attaching a monitor (though, I suspect, I'd need a dummy output, since I don't believe the Windows installer boots if there's no output attached..)

Remilia: yes, and if you'd had to setup FreeBSD that often I suspect you'd have a bit more familiarity with the process.

yeah but even without that it takes me less time than with Windows

Its not just setup. Its so many things. I can repeat if Remilia is new : Brave browser, netflix, wacom pen, bluetooth versions support, less programming app/packages eg nodejs, this list goes onn.

Remilia: Ah, I misread, sorry!

when I say set up I mean everything needed for my tasks

oh ok

configuring Windows for my needs takes two days

because I keep hitting stuff that works the wrong way that I forgot to fix, etc.

no script/batch files help in windows?

I stick to regular Firefox for reasons beyond the scope of this conversation, and have nothing to do with Netflix since I don't like paying subscription fees for things I don't use often - but bluetoth, wacom pen, and third-party software aren't things I've really run into issues with.

Microsoft has routinely made it more difficult to automate the processes.

It used to be that you could slipstream updates onto the installer media, with a little care - but now you basically have to go the WSUS route.

IME, bluetooth devices were easier to pair with a laptop running Linux

There's also a _severe_ lack of a proper package manager, out of the box (yes, winget is useful in one sense, since it's what's used to install Chocolatey..).

debdrup: many in this channel reported bluetooth issues in my presence. It's ok about your feedback but many people do want netflix, brave etc. As you said, we should not assume

Beladona: it's one thing to want something, but another thing to expect it; at the end of the day, FreeBSD exists because of volunteer effort, not because a company is investing money into it - that's the fundamental difference between FreeBSD and the majority of Linux distributions that it gets periodically compared with.

I respect the free source community!

I understand!

I'm sure you do, but it's a bit hard to read that through all the expectations you're outlining on what you think should work, is my point.

Right: I wish something could be done regardless of the situation you wrote (free/volunteer etc). I am not sure what but this OS deserves more.

also to be honest modern wifi and bluetooth situation is terrible more often than not

especially wifi

At the end of the day, FreeBSD _isn't_ a desktop OS. It's an OS in the older Unix-like style, where you're given a compiler, the source code, and instructions on how to build the OS itself. From there, you get to use it as a building block to do whatever you want with it.

Those building blocks can take the shape of a desktop OS, but it's still up to the user to do a non-insignificant amount of work.

I don't think the FreeBSD project or community is going to do away with that notion any time soon.

I just lost hope that if I would buy something (bluetooth), or install something (nodejs, clojure/rust), I might not find the hardware support of software version supported. Tangible example: No support for puppeteer (a node lib). So no point in using the OS despite it is my favourite

Anyway, I'll get off my two-step ladder and leave the floor to someone else.

many other examples. No NVM. (node version manager).

debdrup: yes. BUt I must say I had a lovely time in Freebsd

debdrup: hey thanks for all the support !

:)

Beladona: a lot of modern bluetooth and wifi hardware are black boxes with proprietary drivers and no documentation

ya..

If one wants a pre-configured FreeBSD desktop OS, use GhostBSD. Even though I'm "power user", I do like that FreeBSD comes with a solid base to build off of in DIY way, be it for desktop, server, or embedded.

ok but still the issues I listed will be there

got to go

Remilia: Hi again, missed what have been talking here recently ^^ anything new regarding your issue?

nope

I have no idea how to investigate it properly and no one else probably runs 10+ jails with relatively low amounts of requests per second

Remilia: Do you have unnecessary periodics disabled in each your jail by a periodic.conf file for each?

yes

I've seen they run lots of pseudo stuff causing high CPU load, in the past.

for nothing.

Remilia: I have this for each Jail of mine: pastebin.mozilla.org/ssuv1WMS

to be honest I am not sure how to explain that I had zero issues even without disabling periodic scripts back before I upgraded to 14.0

Title: Mozilla Community Pastebin/ssuv1WMS (Bash)

and how to explain that yes, I have them disabled since yesterday for testing purposes

powerd question here, i am noticing that on one of my machines.. powerd just "dies" freebsd 13.2-p4 and when i run powerd -v this is the output i get: bsd.to/Zi2V

Title: dpaste/Zi2V (Plain Text)

my question is, would it "hurt" to "allow" the CPU to go up to 3000mhz.. this is a pi4

considering that CPU load never spiked for more than 2-3 minutes prior to the upgrade, and that I upgraded using the recommended method, with the latest freebsd-update, from 13.2-RELEASE-p5

just seems like i am "limiting" it if it wants to burst higher

voy4g3r2: if it'll clock that high up. You're gonna need some major improvements to its cooling.

tercaL: there should be a way to disable all, and only enable the ones you want

It'll throttle down if it can't dissipate the heat it generates.

magnahelix: i got a fan and heat sink on it and it is in a basement area where quite cool.. but better to be safe than sorry

i was just not expecting it to jump that high up.. since it is rated for 1500mhz

2000mhz is entirely possible with that pi4.

Maybe even 2.1 -2.2 ghz.

meena: How?

Remilia: i think you have enough circumstantial evidence to file a bug (and assign it to mjg@ ;)

Remilia: and for each your jail, in their rc.conf files: cron_flags="$cron_flags -J 15"

I am not sure what would that help with

tercaL: i don't know, but going from implicit to explicit would be useful

Remilia: Preventing cron jobs to start at the exact time for all jails

tercaL: she's got like 8 jails, not 80 or 800

tercaL: allow me to explain, I have been running the same setup on FreeBSD 12.x and 13.x for years, with no signs of trouble

I probably forgot to mention this before

magnahelix: i would just like xfce4 to NOT tell me powerd died unexpectedly.. thi sis a dev box for a project.

tercaL: two weeks ago: i.koumakan.jp/2023-11-26/1701008382.png this week: i.koumakan.jp/2023-11-26/1701008410.png

i was going to install libreoffice but gnumeric is causing enough headaches.. the only app i am using on xfce4

the difference between these two graphs: FreeBSD 13.2-RELEASE-p5 vs FreeBSD 14.0-RELEASE

nothing else

Remilia: please open a bug report. just to have all the info tracked, and everything we've tried to exclude any hypothesis

I'd open a bug report.

meena: I guess I will have to :\

core developers are just people

most of them, anyway, as far as i know. I've only met one in person, so this is a big extrapolation

meena: imagine getting a reply from d.e.j. under your report

I dunno who that is

well xfce4, powerd do not like each other.. boo

i guess time to find some other wm to rdp into

wow hard core unresponsive.. but xfce4 looks soo pretty

voy4g3r2: hah

what's powerd doing to it?

there's also powerdxx in ports. it might be able to do better

meena: i have no root cause, i load xfce4 i try to load gnumeric and it just "hangs" then i get powerd unexpectedly crashes

I forgot about this, don't know where I got it originally, but I have this in my notes and forgot how useful it is...

comm -23 <(sysctl -T -a -N | sort) <(sysctl -W -a -N | sort)

That command is labeled as: "List tunables that need to be set in /boot/loader.conf vs /etc/sysctl.conf"

i will borrow that :)

_xor: cool

-sh: Syntax error: "(" unexpected (expecting word)

things never work in FreeBSD!

pstef: less symbols more words please -- signed, /bin/sh

Title: "Weird Al" Yankovic - grammar lesson - YouTube

pstef: <( is a bashism, not implemented in FreeBSD /bin/sh AFAIK

bash comm -23 <(sysctl -T -a -N | sort) <(sysctl -W -a -N | sort)

I'm spoiled with having a history command, I tried sh, but it's missing that killer feature.

yes I'm aware it does create a history file

I'd be massively unproductive if I didn't have my history XD

thedaemon: bash -c 'comm -23 <(sysctl -T -a -N | sort) <(sysctl -W -a -N | sort)'

so if it does create the file then what is it missing?

otherwise the current shell still tries to expand the <()

ahh :)

pstef, you have to cat the history file to view the history

there is no "history" command in the shell

cat ~/.history

there is fc -l and I think recently we added a "history" alias for it

what's the "man" command to find the help for fc?

I forgot XD

man fc brings up builtin(1)

man sh

there is another way

ahh it's a bash feature

"help fc"

something else sh is missing

lol

pstef, thanks for the tip!! :)

Title: Oh My Zsh - a delightful & open source framework for Zsh

I learn new things every day

i really need to get my configs in a git repo... this is becoming painful

do it, I find it very helpful

hrm.. interesting that xfce was working, even though i didn't have an /etc/fstab for proc /proc procfs rw 0 0

why wouldn't it work?

pstef: not 100% sure, just hte handbook says it needs to be there and i am trying to figure out why powerd just crashes .. when it is loaded

can you show the handbook page where you see that?

Title: Chapter 8. Desktop Environments | FreeBSD Documentation Portal

and so far so good.. i just did a reboot for sanity of test.. but it is actually "working" but will keep playing

all i have running, minus xfcs, is gnumeric to do some excel stuff

very interesting, because I've been running xfce for a long time without procfs mounted

procfs is optional on FreeBSD

but is it optional for XFCE is my question

the way that i read that section, it appears to be required.. at least it is implied

well it does say require, so one would say it is

and since i added it, no powerd crashes.. so far

I cannot find information about it, but it appears to be a linuxism

man 5 procfs

yes, it is.. dbus requires it to work.. i do know that

well at least it was "always" there from that world.

I have it in my /etc/fstab and I don't even use xfce

I think it was an Irixism

probably for dbus

It was in Unix 8th edition XD

sysctl is the new way in FreeBSD and OpenBSD killed /proc years ago

makes sense to me, explains why i had to add it to /etc/fstab but so far xfce is much much more responsive

nice :)

i think xfce only needs /proc for a few features to work (like powerd)

I need to get powerd to work speaking of that

and smartd

CueXXIII: which explains why powerd kept crashing and since i enabled that /etc/fstab "feature" it has yet to do it :)

After running a regular poudriere update, Nextcloud fails to update. I see `ld-elf.so.1: /usr/local/lib/php/20210902/apcu.so: Undefined symbol "php_pcre2_match_data_create_from_pattern"`

How do I figure out a) which port installed /usr/local/lib/php/20210902/apcu.so and b) how to fix the problem

Schamschula: pkg which /usr/local/lib/php/20210902/apcu.so

The apcu.so extension is probably compiled for a different PHP version than you're using

dh: It needs a revision bump, after the update to php 8.1.26

From which version?

8.1.25

i.e. [7/46] Upgrading php81 from 8.1.25 to 8.1.26...

Hm, that's rather minor difference, usually PHP extensions are not that picky

Hm, does your php have pcre module?

php -m | grep pcre

You would thing

s/thing/think/

Never assume, always check & verify

It gives me pcre

ok that's not working

Someone here who can verify a statement / question about a zfs pool upgrade before I do it?

I have read: freebsd.org/releases/14.0R/relnotes

Title: FreeBSD 14.0-RELEASE Release Notes | The FreeBSD Project

Based on those instructions i have "mount_msdosfs /dev/nda0p1 /boot/efi" and copied a file "cp /boot/loader.efi /boot/efi/EFI/BOOT/BOOTX64.efi"

thorre: double-check /boot/efi/EFI/FreeBSD

Is that all that I need to do before I upgrade my ZFS pools?

there may be a loader.efi there that's actuallyused instead depending on your boot config

[0]helio~ # find /boot/efi

/boot/efi

/boot/efi/EFI

/boot/efi/EFI/BOOT

/boot/efi/EFI/BOOT/BOOTX64.efi

[0]helio~ #

kevans: seems like BOOTX64.efi is the only file

Watch out for the block cloning bug if you're upgrading your pool.

psychonate: Isn't that disabled by default?

on release yes

kevans: I am on release.

and cool re: only \efi\boot; that's a bit of an older setup but it's valid

So, the file that I copied to /boot/efi. Is that all I need to do before upgrading my pools?

should be, yes

kevans: THank you for verifying

thorre: I copied it to /boot/efi/freebsd/loader.efi

that'll make sure the copy of loader matches the feature set enabled on the pool, which should be sufficient once we've made it all the way to a release (in general)

dh: their setup doesn't use this one

If I remember correctly, I did that AFTER not paying attention to zpool upgrade issues

systems installed before 13.0, or maybe even 12.0, just used the removable media default in \EFI\Boot; we've since added \EFI\FreeBSD\loader.efi and create appropriate efibootmgr to establish it

Title: Paste.to

yeah that's fine

I think mine has been upgraded for the past 4 or 5 years so that makes it to old for production but just right for home use.

OK, will upgrade and reboot, wish me luck.

good luc

That was quick.

Yes, it made it!

zpool status is all good

not that quick compared to my home router (a 6W AMD Jaguar system)

time to give the pools a good scrub

woot

you might also be interested in setting vfs.zfs.dmu_offset_next_sync=0, but the reports we've seen so far on FreeBSD seem to mostly be in part of the build system for src

Think I missed to copy the efi-file when I upgraded to 12-RELEASE. Since the machine I run is headless that was a real dumpsterfire.

i went to upgrade a machine recently that's running along -CURRENT and there was some ACPI breakage for a period of time. I knew about it and still forgot to update my loader.efi. the machine was even worse because it was headless, no serial option and there was only one display in my house that it could drive without panicking at boot

and that display is a 55" or so TV on the other side of the house

Upgrading my machine at home is always a bit traumatic. Never have that feeling when doing things like that at work. I guess that my machine at home is one of one. If that breaks internet is gone and the wife and kids go mad.

What does vfs.zfs.dmu_offset_next_sync=0 do?

It currently has the value of 1

helps mitigate openzfs/zfs #15526 until patch is out and ready at the cost of some holes being expanded iirc

Title: some copied files are corrupted (chunks replaced by zeros) · Issue #15526 · openzfs/zfs · GitHub

15526 – [NEW PORT] security/pgpgpg: a wrapper for GnuPG to emulate PGP 2.6 bugs.freebsd.org/bugzilla/show_bug.cgi?id=15526

it's a conservative option, if you want to play it safe

Does not look good

it's not a common corruption

vfs.zfs.dmu_offset_next_sync: 1 -> 0

Have changed it now, must read up a little more on what it actually does

When i run "zpool get all <poolname> | grep feature@" it looks like block cloning is enabled

Is that something that I should disable for now?

yeah, but there's a sysctl to effectively disable it that should be flipped on (disabled) in 14.0

vfs.zfs.bclone_enabled

=0

s/on/off/ :-)

sysctl outputs "vfs.zfs.bclone_enabled: 0"

zpool states

zroot feature@block_cloning enabled local

yeah, 0 = disabled

I am no ZFS expert but to the best of my knowledge the block cloning is part of the copy-on-write design of zfs.

Guess that I have to scrub weekly instead of monthly now.

Is there any information on how common these corruptions are?

Hmmm, could it be that the zfs-pool supports block_cloning but the block_cloning is disabled in runtime via vfs.zfs.bclone_enabled=0 ?

thorre: there is no "root cause" yet, people have been able to confirm it happens.. the exact reason why.. still unknwon

Title: ZFS - Don’t enable block cloning just yet. | The FreeBSD Forums

I saw that article.

Guess I need to pay extra attention to my scrub results.

Since it only is a home server there is not much activity on the disks. But for production this is some extra excitement.

and i have to write that down myself. "playing" with zfs and figure it is important.. to scrub

Scrubing is like flossing ones teeth. It keeps the rot away.

voy4g3r2: I don't think that is a correct characterization of the situation

there's at least two patches in flight to address the problem (when combined)

Indeed: I locally revision bumped php81-pecl-APCu, and now Nextcloud is working again.

I think I accidently removed some dot .files and cannot login to awesomevm. I am in shell now with no browser. Does anyone knows what the config file is?

its was some dot file in home dir

Beladona: This might be useful: epsi-rns.github.io/desktop/2019/06/15/awesome-overview.html

Title: Awesome WM - Overview

mason: no browser : )

Beladona: It works in elinks.

I think it was some file having some GUI entry to run awesome on `startx`. That file was a .somefile in my home dir that was deleted.

and w3m

.xinitrc ?

Beladona: Then maybe "exec awesome" in .xsession

or whatever the binary's name is, exec'd from there

.xinitrc

echo "exec awesome" > .xinitrc

mason: yes. but are you sure its .xsession and not .xinitrc?

You use startx, and not a desktop manager?

Beladona: Yeah.

Unless you use startx.

I do `startx` always

Then it's .xinitrc

ok thanks. doing

it was .xinitrc and all that worked. I am back in UI. Thanks!

enjoy :)

Having lynx web browser available on a headless box can be a life saver, IMO. IIRC elinks and w3m work best from an xterm window on a running desktop, which usually ain't gonna be available on headless servers.

unixman_home, What issues did you get with elinks & w3m when run on a headless box?

parv, I don't recall. It's been a *long* time since I was having to pick a CLI web browser. TBH whatever was the problem back then may no longer be an issue. I've just been using lynx for years.

unixman_home, Ah. Thanks. Did you happen to try links2?

parv, I think I have. It's not like I have used CLI web browsers daily, weekly, or even monthly. It's been at least a year or more since I last needed one. It's one of those, when I *need* it I use it, things for me. I don't often find myself in that position.

The last time I recall using lynx I was in the DC, with a crash cart keyboard / mouse connected to a headless box that had lost M&C connectivity. I used lynx to connect to a vendor site and download a patch for their software running on that box. After that, I worked with network engineering to fix the M&C connection so the next poor admin would not have to go to the DC for patching that box.

unixman_home, What is "M&C", "monitor & control"?

Pretty much. I think we call it Management & Control, but basically it is the remote management interface built into a lot of server hardware that we connect to a dedicated network for that.

Does anyone here use the enlightenment desktop? I have a question: I'm on a fresh install, just started e and after the 'wizard' setup thingy, all I get is a black screen, it is "as if" the backlight went to 0

also is it really bad to mix ports and pkg installed programs?

I remember at one time it was not really a good idea, then it was considered 'ok' but now?

doing some tests with sockets, using old examples. bpa.st/2DIQ I am struggling to understand how to extract the IP from struct hostent . Any help is welcomed.

Title: View paste 2DIQ

back

AumShivaya: it's fine to mix ports and packages

uskerine: what is h_addr/ line 45 showing you ?

uskerine: You need to use inet_ntoa()

Haven't used E16/e17 in a long time. Sorry. mixing ports and packages is fine like rtprio stated. packages are just precompiled ports.

.. and you need to do some casting - stackoverflow.com/questions/15670713/using-gethostbyname

Title: c - Using gethostbyname - Stack Overflow

AumShivaya, Mixing ports & packages may not work in some cases if you would use non-default options for port building due to differing dependencies

"it's fine to mix ports and packages" comes with a fat caveat

uskerine, A snippet from one of my old programs that might help a little: bsd.to/ouWs

Title: dpaste/ouWs (C)

rtprio dh bpa.st/I4SQ

Title: View paste I4SQ

uskerine, inet_ntoa() returns a string not an unsigned.

so when I use p_server->h_addr as a struct in_addr pointer I get a decimal value which maches the IP (although it is read the opposite way)

What you say is the opposite way is the difference between big-endian and little-endian.

Most systems are little-endian these days. But networking came about when they were big-endian.

I'll go out on a limb and say all network protocols are serialized as big-endian bit streams. But Intel x86 systems are little-endian.

So data always needs to be converted from host ordering to network ordering on the way from the host to the network wire.

And all data needs to be converted from network ordering (bigendian) to host ordering (littleendian) on the way in from the wire to the host.

thanks :)

If you happen to be working on a natively big-endian system then the byteorder(3) macros/functions do nothing. But on a little endian system they flip the bits around.

for some reason I get a segmentation fault while trying to apply inet_ntoa

Can you use the returned char * directly as a parameter of printf using %s ?

Turn on all compiler warnings, re-read the man pages for every function, double check all of the types.

rwp: pretty much all old-school systems that are still around are now Bi-Endian, except for IBM's big iron

uskerine, Yes. I am doing this: sprintf(peername,"%s",inet_ntoa(peeraddr->sin_addr));

meena, That's why I said I was going out on a limb. But general guidelines are useful for newcomers to go by. There will always be exceptions.

well i guess you could switch your aarch64 systems to Big Endian, if you enjoy pain

build your own exceptions!

I am sure you are familiar meena with the classic "ON HOLY WARS AND A PLEA FOR PEACE" by Danny Cohen but uskerine should read it rfc-editor.org/ien/ien137.txt

rwp: I probably am and it's been too long, but either way, I don't remember (much these days)

It's a fun read and if dealing with endianness such as the byteorder(3) macros then still applicable too. But otherwise the machines it references are all museum pieces now. Machines prior to 1980.

A header was missing

and inet_ntoa was going ballistic, I reduced the problem to just getting the IP address of www.yahoo.com

normally I do not cease until I get no warnings, but this time I was not reading them

Missing header files will mean missing prototypes. If warnings are enabled then these will be caught at that point. And the man pages always list what header files are needed.

I am aware on the little endian big endian, and that network is big endian

But don't go crazy adding every header file to every file thinking to avoid not having some needed header file. That way is considered bad as it slows things down.

I once had a complain from a junior because my software was assuming that the architecture was little endian. I was wondering where he wanted to compile that thing in 2023 as everything that you put in a data center is little endian.

Thompson apparently had such an issue with it at Google that in Go-lang it is an error to include a header that is not used in the compilation unit.

Yes go-lang is very strict on that end, same for unused variables

I normally like to know which header is for what function, but with sockets it seems that you need several headers at once

Networking will always require several headers all together. True.

I am trying to make peaces with sockets, and learn them in a reasonable proper way

Excellent! It's useful.

rwp: to be fair, on Plan-9, you can't include header files more than once, and that makes for a lot, and i mean, a lot faster compile times

So it's not just unused crap, it's also multiple incudes of the same thing

Agreed. I have worked on projects where there were massive lists of includes and one of the problems is that it kills incremental builds. Change anything and everything becomes out of date.