ack

lw, you can use the fetch option in poudriere. this will cause it to download binary packages if it would build the same thing

jbo: i think that might not work well because i have a load of global options, but i should probably make an empty options dir for testport anyway

oh man

i got a filesystem error

and recovering was impossible

had to reinstall all

that seems unlikely :)

what happened?

lw, cannot set property for 'storage/backups': 'mountpoint' cannot be set on dataset in a non-global zone

journal timestamp did not match the filesystems last mount date

jbo: are you in a jail?

lw, nope

that's odd

and repairing and rebooting didn't change anything

yep

did the filesystem somehow end up with jailed= set anyway?

oh, that could be it, because I did pass that to a jail previously

storage/backups jailed on local

:)))

thanks!

this reminds me to run syncoid again since my backup host got some more memory

lw, I am still not sure whether it's nicer to mount the dataset on the host and then nullfs mount it into the jail or whether I should give the jail the dataset

like generally, nothing to do with syncoid

lw, setting mountpoint works now - thanks for the hint!

i tried both and found mounting it in the jail is very awkward

you mean nullfs mounting it?

the jail won't unmount it on shutdown, i had to edit /etc/rc.d/zfs to do this or the jail was stuck in dying forever waiting for the filesystem to unmount

sorry, i mean mounting it via jailed=

aye

i went back to nullfs which is more annoying but seems to work more reliably

I guess if the jail doesn't need zfs features it's fine.

I mean if the host were to be compromised you're screwed anyway. it's not like passing the dataset to jail would help there

if this annoys me enough i might investigate it more, it definitely seems like it *should* work because the jail.conf options exist, but it doesn't, and if i'm doing something wrong there's no documentation to refer to...

i tried using testport -b latest with default options and it only fetched 5 packages out of 199 :-(

current repos are probably not fun

also, what helps is ctrl+^ and then restart :p

(not kidding)

it usually fetches more on the 2nd run

(if it can)

huh, now it fetched >100 packages. that's weird

thanks

:D

the ultimate pro gamer move

spawn camping for ports

I can have multiple mount += ; lines in my jail config, right?

ah, I guess I need to create the destination mount dir in the jail

i'm not sure, i always use mount.fstab. i vaguely remember trying to use multiple mount options didn't do what i expected

hmmm

I added a 2nd entry and it tells me that "no such file or directory" on the jail dest path

lw still getting screwed over by the broken pipe

lw, ping

jbo: i think this must be caused by an earlier error, but if it doesn't show any other output... is there a verbose mode?

hm: ===> mediaelch-qt5-2.10.6 cannot be installed: unknown USE_QT component 'qmake_build'.

but the handbook says to use qmake_build buildtools_build

it's qmake:build afaik

oh, it's buildtools:build qmake:build

grep to the rescue

USE_QT= buildtools:build core gui qmake:build widgets

yeah

anyone know what "private" means in autobridge_bridge0="..."? or where i can look for the docs?

lw, jimsalterjrs/sanoid #427

Title: Syncoid breaks pipe on Proxmox · Issue #427 · jimsalterjrs/sanoid · GitHub

alepzi: i would guess that 'private' flag documented in ifconfig(8)

jbo: hah. guess i never noticed that since my backup host is linux

there we go, thx lw

jbo@hq:~ % echo $SHELL

lw, no more broken pipes after the chsh .__.

csh.whynot strikes again

yeah screw that. default shell should be sh in my opinion :p

although i wonder how mbuffer could possibly care what the shell is... does it use $SHELL to invoke commands?

it is sh :-d

old host

wow, finally a successfull backup

WARN: mbuffer not available on target ssh:-S /tmp/syncoid-jbo⊙111 jbo⊙111 - sync will continue without target buffering.

I guess I'm fine with that

not sure why you'd want to mbuffer on a ZFS system anyway?

i think it's used for the progress bar in interactive mode, not sure it does anyone useful in cron

s/anyone/anything

my progress bars don't reach 100%

lw, the first time you recommended sanoid/syncoid to me you did mention that you can do some housekeeping on the dest host (i.e. automatically removing old snapshots). how did that work again?

jbo: i haven't actually set this up yet but the idea was to run sanoid on the backup host with the same configuration as the source host except autoprune=yes, autosnap=no, which means it'll delete snapshots without creating new ones

so it deletes the old sanoid snapshosts that exist on the source

reasonable

obviously only applies if you're using sanoid on the source, if you use something else i guess you have to do it manually

I don't really like that sanoids calls the first snapshot "clean"

why not "backup_start" or "backup_origin" or whatever

i don't think it does, that's poudriere's snapshot

"clean" is just so... noh.

oh

never mind then :p

zroot/poudriere/jails/main@clean 240K - 2.52G -

yeah that mekse sense - sorry

I was using a poudriere dataset to test sanoid/syncoid because it had hierachic datasets as well as neither to much nor to little data for some testing

if i want to combine the contents of all files in a dir, is the syntax cat * > combined.txt?

seems to create a huge file for some reason

alepzi: yes, but perhaps your shell handles the output redirection before it expands the glob, so combined.txt is included in the list of input files... i just tested /bin/sh and it doesn't work this way though

lw, now trying to figure out whether it's worth running sanoid under a separate user rather than just root

ya i'm using the root shell. csh or something

can't install bash because i messed up the network somehow

alepzi: yes, csh does this the bad way. don't use csh, use /bin/sh. if you really want to use csh, use ../combined.txt or something

there's less reason to use csh in base nowadays since /bin/sh has learnt some new features like history editing

aha

jbo: on the source host it's hard to imagine getting much benefit from that... you'd basically have to delegate the entire pool to it, right?

since updating from 13.2-R to 14.0-R: /usr/src,/usr/ports fail to fetch saying "fatal: not a git repository"

lw, indeed, that's what I'm considering

concussious: you're running 'git fetch' in /usr/src? does /usr/src/.git exist?

lw, still feel like using the same user as I use for syncoid on the source host but then a doas.conf rule

with nopass

jbo: so the user could zfs recv a filesystem containing a setuid /bin/sh or something and get root

it doesn't in either directory

(or can you prevent it from running recv?)

lw, well, as you know I was very pedantic about having one user per host on the dest host. that also means that I can have the user's shell set to /bin/sh :)

however, .gitignore exists in /usr/src

concussious: did it work before? how did you upgrade?

jbo: i'm just thinking that effectively giving root access to a non-root user might decrease security as non-root accounts tend to be less protected. maybe that's just a theoretical problem though

it did work before. both directories were created with git following the handbook. updated using freebsd-update upgrade

concussious: do you remember if freebsd-update upgrade updated the src set?

lw: i don

hmm, i wonder if it detected src installed and replaced your git checkout with the src.txz set... but i don't think it should do that

check 'Components' in /etc/freebsd-update.conf and remove 'src' perhaps, but i suppose you won't be able to test that until 14.1 is out. in the mean time you might need to rm -rf /usr/src and do another checkout (backup your kernel confs or whatever)

lw: nice suspicion! src is selected in freebsd-update.conf

my computer connects to my router. my computer also runs a bridge/tap network for bhyve vm guests. termbin.com/30uc anything wrong stand out? it *was* working, now i can't even ping the router ip

jbo: does poudriere testport save the package somewhere so i can install it and check it actually works? /poudriere//data/packages/main-development/All/mediaelch* doesn't exist...

lw, well, hence I would just add a doas rule which ONLY allows it to run 'sanoid' without entering password

lw, no, you have to run poudriere-bulk and then grab it AFAIK

if you setup ccache that will go quick :)

lw, there is an interactive mode for poudriere-testport which you can enter with the -i switch. that drops you into a shell of the jail AFTER install but BEFORE remove. not sure how helpful that is with GUI stuff tho. it's great to do dependency investigations and stuff tho

alepzi: does the ifconfig -a output look reasonable? what does ping say?

it's a bit unusual to use _alias0 for the primary ip but i think that should still work

i've never used autobridge though

i do not understand how you're meant to fit a meaningful commit summary into the 50 characters that git commit wants you to use: le-fay.org/tmp/30d/Screenshot.from.…024-01-13.at.02_25_29.478505553.png

how can i bind the volume keys on my thinkpad x230?

lw omg figured it out

i was missing an ifconfig_igb0="up"

:)

the best way to retain what was learned is from mistakes :)

meena: cb350ba7bf7ca7c4cb97ed2c20ab45af60382cfb looks like it should fix the krb5 issue but that was committed on dec 6, and it failed for me on dec 27...

maybe i should re-test

lw, when using sudo -u does that "point" towards the SSH priv key in the sudo'd users home dir?

jbo: i think that might depend on your sudo config? at least for me it does set $HOME so it should pick up the target user ssh keys

lw, I'm starting to wonder why I am bothering with this. I could just use root for everything on the src host and still have a per-machine user on the dest .__.

i think i'm going to try running poudriere with -J1 and set ALLOW_MAKE_JOBS=yes... then just send some PRs for any broken packages

(this comment inspired by me sitting here for 45 minutes waiting for llvm15 to build with 4 make jobs)

lw, if you run poudriere inside of a jail you can easily do ALLOW_MAKE_JOBS and then using jail resource control to limit CPU to like 90% or whatever so you can keep doing something else :)

that's an interesting idea

i didn't know poudriere worked in jail

i haven't played with rctl at all yet, i've been meaning to have a look at it

when you run poudriere in the jail, you have to give it permission to spawn it's own jails and stuff

and pass through the ZFS dataset rather than nullfs

lw, github.com/freebsd/poudriere/wiki/poudriere_in_jail

Title: poudriere_in_jail · freebsd/poudriere Wiki · GitHub

why does git have to be so difficult

lw, that is an "old guide". the patching is AFAIK not needed anymore

voy4g3r2, what do you perceive as difficult?

let me rephrase: what are you struggling with? :)

so i create a git --bare init

because i have a directory of stuff..

get it all commited and what the heck is up with this damn master main crap

just relax - people being people.

do a git init on directory, then update my .gitignore (to minimize stuff)

it's just a branch name tho. you can also not have a branch named master or main

git commit -m "blah"

you probably forgot to git add first

git push

i am like awesome, everything is good

then i got my bastille jail.. try to get the dang repository

says no main exists.

you can always have a look at mecurial/hg

at first, old habit was master, then switched to main (i honestly do not care either way)

it's an "easier git" apparently

that is all.. but at least i can now isolate this manual pages update work to a jail

instead of "soiling" my host

jbo: but basically just a rant.. that is all

but back at it.. beck - morning phase to the rescue

voy4g3r2: what are you trying to do?

git branches are not that complicated but it can be a bit of an adjustment if you're coming from another vcs like svn

jbo: thanks

jbo: looks like dan.langille.org/2019/10/23/moving-…poudriere-from-the-host-into-a-jail might be more up to date (linked at the top of that post)

Title: Moving poudriere from the host into a jail – Dan Langille's Other Diary

ah nice - thanks to dvl :)

lw: i am just trying to move a project from one machine to another machine (a jail) to isolate work

voy4g3r2, rsync?

it is just more of a frustration and i think i got it going

hm... i probably don't need to back up 54GB of /src/obj. also, why is my /src/obj 54GB?

oh, because i build a make release

also 12GB of /var/crash might be unnecessary

this is why i never have any disk space, everything is conspiring against me to use it all

it's amazing how much faster MediaElch is on FreeBSD with NFS compared to macOS with either NFS or SMB. on macOS, it was 'start a file scan then get a coffee'. FreeBSD scans the same files in under a minute.

(for xen+pygrub/pvgrub)

I have currently linux on one drive, and freebsd on another

I switch between them in the bios

uefi should make this easier

UEFI doesn't really make this easier because it depends on your BIOS providing a way to change boot device easily

well it does not provide a menu

have to click here switch that

and exit and save

reboot

not optimal

that's annoying. the UEFI systems i have here both (i think) provide a boot menu on F11

oh? perhaps there is but I was not aware of this

(one MSI, one ASRock)

ok

I shall brb

was going to say i'm not sure about the ASRock because i haven't connected a monitor for ages, hope it works out for them :-)

f11 it is

any luck?

oh good

yeah, be nice to have an option whereby it just came up without pressing f11

possible/

?

i don't know, consumer BIOSs tend to be pretty limited when it comes to this sort of thing. it might be worth checking for an option like that but i don't remember seeing one here

but yes, this saves time anyway

although, my MSI boards to a UEFI command line if it doesn't find a boot device, so i guess you could do that and then load the loader by typing the command... but that sounds really annoying

s/boards to/board drops to/

I was thinking FreeBSD had a chainloader that could load linux or freebsd?

i've never noticed anything like that but i've also never looked for it. it might be easier to do it in grub

i don't even know if you can chainload EFI loaders...

didn't FreeBSD use a chainloader at some point?

It does not now, I do not think

I think at one point there was 2 loaders tho

on legacy BIOS it uses a bootblock that loads the real loader, which you could say is a chain loader

yeah, probably that

this email confuses me every time i see it: 1 Jan 13 portscout⊙Fo ( 31) FreeBSD ports you maintain which are out of date

no, portscout, i do not maintain libxml2

ok next question, why is /src (without /src/obj) 32GB

jbo: 301MiB 0:00:08 [33.8MiB/s] [===============================================================================================================================] 113%

jbo: percentages are hard?

(syncoid)

then one finished at 17% 720 B 0:00:00 [ 296KiB/s] [====================> ] 17%

lw: i have found if you tweak the mbuffer size, it gets "better"

i mostly run it from cron so i'm not too bothered

when i was testing it out BEFORE i got to cron.. i tweaked that option

i wonder why i'm only getting ~10MB/s. maybe increasing the buffer size would help with that

i noticed an increase when playing with that option myself.. i got to 256M buffer as sweet spot

oh, i think it's because wireguard is making my router sad

voy4g3r2: --mbuffer-size=256M?

yup

syncoid --recursive --no-privilege-elevation --mbuffer-size=256M --sshkey=/home/backups/.ssh/id_rsa storage/bastille "${REMOTE_HOST}:storage/backups/bastille"

as an example

i have the --no-privilege-elevation because i did not like running as root

that seems to have helped a little but it's fluctuating between 500kB/s and 60MB/s in a weird way

it may be due to a lot of small files? i am maining moving mp3/flac files and jails

they are quite large.. so it why i went with the 256, i think it defaults to 128

if i recall

ha! bastillebsd defaults to thin jails...thick jail it is..

never understood the point of thin jails, just use dedup

i know everyone is always like "omg dedup is evil" but for jails it's absolutely fine, you don't need 1TB of memory

yeah that is way outside my understanding.. the whole dedup stuff. i just know i want to be able to work on /usr/src and thin jail does not allow it.. and hard drive space is "cheap"

voy4g3r2: if your jails are on zroot/jail, just zfs set dedup=on zroot/jail and enjoy your free disk space. (unless you have huge filesystems under /jail... i just put the OS itself there and mount application storage from elsewhere)

thats good to know.. i have bastill on zroot but i have a huge storage pool with a few terabytes

oh wait.. nah everythign is on storage pool

but noted the suggestion

freaking trailing whitespaces in diffs

=gr

right, i have /data on a separate pool and i mount storage from there to jails when needed, so dedup only applies to the jail's root filesystem

if you try to dedup several TB of storage you will need more memory :-)

32 gig of ram but we shall see

the zroot has 256 gig of space.. which is much lighter

i do have nvm.e drive on my wish list along with a dual port intel nic card.. to get rid of my HORRIBLE router

so i can add another hard drive to the /storage pool :) as i am limited to 4 sata ports

but i think it is time to watch this freebsd fridays talk on bastillebsd.. to increase comfort.

good luck lw on the syncoid things and thanks for tips on dedup

lw: I chainload FreeBSD from my Linux systemd-boot (also works with OpenBSD).

AumShivaShakti: ^

hm... syncoid compresses by default, doesn't it? it claims my windows VM is a 41GB image which seems excessive

oh, default is lzo

topcat001, lw, cool thanks for letting me know it is possible

ugh i really don't want to fix audio/sublime-music to meet the python ports guidelines, they're ridiculous

apparently i have to install the executable as sublime-music-<python version> and then install a symlink from sublime-music but only if this is the default python version

like who would ever bother to install two versions of the same app for different python versions at the same time? no one does this

ok i guess it only says you "SHOULD" do this, i am going to ignore that on the grounds that it makes no sense

jbo: i think all my ports have up to date patches now, lmk if not

jbo: also my client only highlights if you say "lw:", not "lw," ... just in case you wondered if i was ignoring you, i'm not :-)

so if anyone remembers me ranting about my weird zfs issue, since i disabled sanoid hourly snapshots, the problem seems to have gone away

I already build my py-* packages with py311 as default, especially to get the extra performance benefits. By now most python libraries should be compatible 3.11 anyway.

yeah, i'm using 3.11 as default too

i'm not even sure if the pkg builder will build my port, since it requires 3.10, and the default is still 3.9... i guess it'll install 3.10 to build it?

(i mean the pkg.freebsd.org builder)

i'm working on another python app that i'd like to port soon and that requires 3.11...

Once your done, if you would like for me to help out with refactoring/cleanup, hit me up

*you're

hopefully it won't require too much cleanup as there's two developers and one is me, so it's pretty much guaranteed to work on freebsd :-)

No worries

darn it

dxx-rebirth fails to find distfiles

freshports has it not listed for quarterly amd-64 just a -

You could copy the port over from latest and try your luck.

hmm, no idea how

I shall first look for a distfile

aren't the distfiles meant to be mirrored on ftp.freebsd.org?

let me see

slow

...reeBSD/ports/distfiles > less README.txt

Our distfiles cache has been moved to:

which gives a 403

I loved Descent

copy over the amd64 supported port: git checkout main -- games/dxx-rebirth

is this port related to 'X: Rebirth'?

AumShivaShakti: dxx-rebirth distfile fetches fine for me, but the port crashes clang when building

-rw-r--r-- 1 lexi users 2071922 Jan 13 06:39 /src/ports/distfiles/dxx-rebirth-dxx-rebirth-20231232-4efe4c9_GH0.tar.gz

on quarterly?

latest (main)

shit

Assertion failed: (Info.Ctx.hasSameType(BaseIt->getType(), BaseType) && "base class initializers not in expected order"), function HandleConstructorCall, file /src/main/contrib/llvm-project/clang/lib/AST/ExprConstant.cpp, line 6374.

PLEASE submit a bug report to bugs.freebsd.org/submit and include the crash backtrace, preprocessed source, and associated run script.

Title: Log in to FreeBSD Bugzilla

i guess i should do that even though i never heard of this port until 5 minutes ago :-)

well, should probably update src first, let's do that

descent fan?

lw: might be worth exploring the games category dir :)

me? no. is it a descent clone? i thought it might be an X clone

which reminds me i need to see if X4 works on freebsd

lw: current or stable?

FreeBSD 15.0-CURRENT #7 main-n267425-aa1223ac3afc: Sat Jan 6 15:39:48 GMT 2024 lexi⊙ielo:/src/main/sys/amd64/compile/ILYTHIA

I guess I can try building dxx-rebirth on 14/stable

it is descent clone

i have a releng/14.0 system i can test on

what is X4?

although it's still latest ports, not quarterly

AumShivaShakti: single-player Elite-like space sim game from Egosoft

(it's really good if you like that sort of thing)

oh space game, x4 foundations?

yeah, that

that on freeBSD?

looks kind of advanced gaming

i don't know, i haven't got around to trying

they have a Linux port though so i'm hoping it'll work under wine

I tried Freespace, did not like

too slow

Overload is the "latest" Descent clone

it is quite fast and fun

i never got into Descent, too many hours wasted playing Wolf3D and Quake

I found descent more fun

multiplayer

that's me

kenrap: i suspect this is a casuality of the clang 17 upgrade, apparently this broke a lot of ports

That would have to be the case, since it's still building fine for me so far while using poudriere-testport(8)

yeah, i'm going to upgrade to current current (?) and if it's still broken i'll file a PR

weird that my CPU won't go above 52C even with buildworld -j8

if you have the ram for it, try abusing more make jobs (to squeeze in a bit more build performance)

might hurt the GUI interactivity though

Alright, dxx-rebirth built successfully for me

no real difference with -j12, dev.cpu.0.temperature: 50.5C

i think these "3D VCache" CPUs just run quite cool

I'm thinking -j32 as an example

if not, -j64

let's try that, for science

I do it all the time :)

it will take up more ram through, so be aware of that to avoid OOM

6:59AM up 6 days, 15:13, 1 user, load averages: 25.86, 12.37, 7.07

dev.cpu.0.temperature: 47.1C

that looks well cooled

7:00AM up 6 days, 15:14, 1 user, load averages: 51.58, 22.60, 11.31

dev.cpu.0.temperature: 49.1C

topcat001: it does have a 2x 120mm cpu cooler + 5 120mm case fans, but i'm surprised the cpu doesn't take advantage of this to boost more

5800X3D is the wrong CPU for this workload though, 5800X would be better

if i'd known how long i'd be spending compiling src, i would have bought one of those. or maybe a 5900X

Meanwhile my work switched from desktops to laptops for developer build machines :|

heh, these numbers keep getting better, it's still at 49.1C: 7:04AM up 6 days, 15:18, 1 user, load averages: 62.60, 40.99, 21.67

topcat001: that sounds terrible

don't you at least get a distcc host or something?

We have some build machines on AWS controlled via github actions.

Another "improvement"

oh, you work in the Modern Enterprise Desktop Environment(tm)

ultra modern - next step is to only have a watch

I actually asked for a 5800X (or equivalent) desktop for my use but IT does not want to deal with custom configs.

i think -j64 made ULE sad

on the subject of -j, i was very surprised to discover that make installworld -j8 significantly reduces the runtime

what are rules for 'Sponsored by' in commit messages? can i just throw this in if i do something on company time?

lw: out of curiosity, what your ULE threshhold setting?

sysctl kern.sched.preempt_thresh

kern.sched.preempt_thresh: 48

wow, gotcha, thank you

ok, something has changed in the current version of sway that has completely fucked up how i interact with my desktop

this is incredibly annoying

i can't even put windows in the right place now without using the mouse, wtf?

this is why i used twm for the last 20 years, it just stays the same and you know how it works. this modern crap changes everything on every update

darn it, no Descent 3 port

i didn't even know we had this many ports: [main-default] [2024-01-13_07h54m40s] [balancing_pool] Queued: 975 Built: 0 Failed: 0 Skipped: 0 Ignored: 0 Fetched: 0 Tobuild: 975 Time: 00:00:16

maybe i should try hikari instead of sway

lw: I was doing some research on ideal make job counts and I found an interesting stackoverflow answer that resembles with your buildworld experiments: stackoverflow.com/a/18853310

Title: makefile - GNU make: should the number of jobs equal the number of CPU cores in a system? - Stack Overflow

So, thanks for trying those numbers out.

Hi.

I'm trying to configure a bhyve guest on a remote server (over ssh) on FreeBSD 3.2 but I got following error: "kbdcontrol: setting keymap: Inappropriate ioctl for device" for any keyboard selected.

FreeBSD 3.2… hello my 1999! :-D

ferz: I'm surprised FreeBSD 3.2 would even get that far under bhyve. but if you mean 13.2, that's a different issue

lw: that start of a commit message you shared didn't look like a summary line, more like the start of the detailed explanation of the commit, so stuff for line two of the commit message, but yes 50 chars can be a little limiting for the summary

Hi. How can I install Virtualbox guest additions through the CD?

The FreeBSD package is too old. It's 6.1.48 and my host Virtualbox is 7.0.12

That's why I want to use the CD.

which cd?

I'm guessing they mean the ISO instead of CD

I'm guessing that they meant one of the official installers and then me next question would be why they think it contains a more recent version than the repos

I mean VBoxGuestAdditions_7.0.12.iso. I believe it contains a more recent version than the repos because 7.0.12 > 6.1.48

The ISO is the CD. Virtuabox calls it "Guest Additions CD image".

no idea; does it contain a package for freebsd? then pkg add

meator: I realized I need to stop playing "inference" games but maybe this is a valid suggestion to make. Can't you just read-only mount the ISO file and copy the guest additions or whatever you need from it?

nevermind

Is there a way to get to prison from vnet?

I am unsure whether the guest additions actually provide a FreeBSD integration. It that isn't the case, the only two mwthods I see, iso and pkg package, aren't possible.

jbo / lw : re poudriere in a jail - I'm thinking of moving that jail from one host (also used for dev/test/stage freshports) to another host (much less used). A new blog post may be coming. I don't know if I'll get to do that this weekend or not.

alright!

that reminds me that I have to figure out how to move jails between hosts without using fancy tools like cbsd

jbo: zfs send | zfs receive, or syncoid

dvl, I guess it would be ncie to have an archive with the dataset and the jail config :p

jbo: You'll always have the data on the old jail until you delete it.

i am huge anti-fan of jail tools. all they do is hiding what jails are under the layer of scripts for all the stuff. they are handy for lazy bones, my humble opinion.

angry_vincent, that's where I'm heading to right now. I was using cbsd for many, many years. been very happy with it but...

cbsd is such a horrible over-featured monstrosity. sorry, but i stay away with overwhelmingly reach things. ok, make sense for very specific tasks. it's good such thing exist, of course.

I use mkjail (also contribute to it) - all it does is create, upgrade, and update the jails - All the configuration is left to you (e.g. /etc/jail.conf).

I've had problems in the past when a jail manager doesn't keep up with the OS - or does things wrong and I can't actually use the base system feature I want to use.

jbo, Im at this right now. Ending like a script that coordinates the move. Thing is going from 13.2 to 14.0, ezjail to bastille, updating mysql jail 5.6 to 8, and other minor things.

I hope bastille will be good to me as it does ezjail

For me at least ezjail worked perfectly, for years. But now it's not even mentioned in handnbook anymore. So it will probably just face a natural death from now on

Seems like I need to migrate to something new. Is bastille the new ting to use?

thorre, I don't know :p After researching in this subject bastille looked like the natural ezjail substitute for me

i can't talk much. Like how mature (or not mature) is it. etc

Testing it. It looks nice

You can 'ezjail-admin archive' ezjail jails and 'bastille import' the tar.gz. Bastille includes code to deal with this

After doing this I had to, 'rm /usr/local/bastille/jails/"$1"/root/basejail' && 'mkdir /usr/local/bastille/jails/"$1"/root/.tempĺate' && 'chmod 755 /usr/local/bastille/jails/"$1"/root/.tempĺate' && 'touch /usr/local/bastille/jails/"$1"/root/etc/fstab' to have imported jails with same 'state' as fresh created bastille jails

Also had to deal with some weird ezjail symlinks that got lost after importing

Thank you for the information DanDare

np!

cool. i didn't know that

thorre, if you go to test it. Take a look. At least for me for every ezjail that I imported ended with broken symlinks in some situations. I created this script to fiz this. Must be run from root jail dir, after importing. After starting the jail, executed inside the jail (e.g. 'bastille console jail') gist.github.com/Rodrigo-NH/195e2f7213c326d0172d0d03a9aa10cd

Title: unbreak symlinks · GitHub

DanDare: I have made a note of it. I will not do it tonight since I had a few beers now ;-)

:)

VVD: yes, sorry 13.2

meena: yes, sorry. It's 13.2

thorre: heads up, if you do a pkg install of bastillebsd on 14.0 and NOT ports.. the online documentation has a NEWER version of the documentation from what is installed in 14.0

I do not use ports on my server

lw: did you end up changing the compression on the syncoid backups? i saw the lzo .. was curious

voy4g3r2: are you on quarterly?

probably not.

ferz: can you upload the whole log somewhere?

well alright then.. definitely not on quarterly, thanks meena

Can hyper-threading be disabled in FreeBSD? I know the normal way is to disable it vi a BIOS option, but on this system I'm looking at, there's no BIOS option to disable it.

I found machdep.hyperthreading_allowed

And it doesn't work. . .

WTF

CrtxReavr: did you set that in loader.conf and reboot?

meena, did you look at the pastebin link I provided?

CrtxReavr: it's not really clear what I'm seeing

machdep.hyperthreading_allowed is a read-only oid, so yes. . . I used boot.conf to set it and the sysctl command I pasted shows it's set.

IT's a dual core CPU, so HTT is clearly still working.

voy4g3r2: no, i'm not backing up enough data to really care at the moment, like ~180GB in total

what's boot.conf?

CrtxReavr: Intel Core i5-2500K has 4 physical cores and does not support hyperthreading ark.intel.com/content/www/us/en/ark…cessor-6m-cache-up-to-3-70-ghz.html

Title: Intel Core i52500K Processor 6M Cache up to 3.70 GHz Product Specifications

lw, look carefully at the "Features" entry: bpa.st/OTQQ

Title: View paste OTQQ

it doesn't really matter what the features flags say, the CPU doesn't support HT... and it has 4 cores, so if HT was enabled, it would show 8 cores

lw, why would the kernel mis-report the presence of HTT then?

I mean. . . based on observed behavior and that ark.intel link, what you're saying makes sense.

I'm just not accustomed to the reported CPU features being wrong.

i will tell you as soon as the Intel instruction set reference opens, but i suspect HTT just means the instructions are available for the OS to query the HT status

i have a network config with weird behavior when i run 'service netif restart'. termbin.com/2eqq

CrtxReavr: yes, this field means "Max APIC IDs reserved field is Valid", i.e. the OS can read the HTT configuration from the provided HT configuration. it doesn't mean HT is actually enabled

(Table 3-11 in "Instruction set reference, A-L")

A value of 0 for HTT indicates there is only a single logical processor in the package and software should assume only a single APIC ID is reserved. A value of 1 for HTT indiA value of 0 for HTT indicates there is only a single logical processor in the package and software should assume only a single APIC ID is reserved. A value of 1 for HTT indicates the value in CPUID.1.EBX[23:16] (the Maximum number of addressable IDs for logical processors in this package) is valid

What started this whole mess. . . from my experience, *nis OSes perform faster with HTT disabled. . . though Windows supports is very well, I wanted to do some benchmarks both ways on this box. . . I guess it's all for naught.

ok WOW netstat -rn after service netif restart is missing the line "default 10.1.1.1 ugs bridge0"

but /etc/rc.conf has defaultrouter="10.1.1.1" in it so wtf?

so the other day i was having these weird problems with poudriere builds using load of swap even though there was plenty of free memory... i switched to -J1 (with MAKE_JOBS_NUMBER=8) and that seems to have fixed the problem. unless it was fixed by upgrading yesterday...

i tried following it with service routing restart. netstat -rn now has the missing line, yet ping still fails

fwiw when i run 'service routing restart' there's a bunch of output but 1 line is "route: route has not been found"

lw: what other settings do you have in poudriere?

meena: le-fay.org/tmp/30d/AFlMyf.txt

the current build has MAKE_JOBS_NUMBER=8, once it's done i might rerun it with PARALLEL_JOBS=2 and MAKE_JOBS_NUMBER=4 and see if that's faster without causing the previous problem

if u crash with 4 u will crash even with 2

use single cpu build

nothing is crashing, it's just using an excessive amount of swap space. a single-cpu build would make no sense, that would take weeks

this build is going fine, the problem is -J1 means you suffer a lot from overhead of the extracting/packaging steps

weird network behavior after running 'service netif restart' is a freebsd bug or? termbin.com/sgm1b

lw: I'm still appreciative of your experiments yesterday since you woke me out of a relatively long lasting misinformation I was given

alepzi-: did you have an existing default route?

ya it worked just fine before the service netif restart

(lmao)

doesn't look like route commands in netif

so what's going wrong?

service routing restart

ya i tried that. the route then shows back up, but ping doesn't work still

that's why i'm wondering if this is a freebsd bug

oh, i just noticed all your wacky bridges

i don't really trust netif restart, i either use ifconfig by hand, or just reboot if i want to check rc.conf is right

there's a lot of stuff involved in setting up the network config and netif only does some of it

why do you have an alias with the ip, why isn't that on ifconfig_bridge0 ?

is there a higher level service i can restart?

don't remember why

does that look kinda shitty?

alepzi-: kill -1 1

just type it or?

it seems unnecessary and broken

i typed the kill cmd

i'll move the ip

ok i made it ifconfig_bridge0="inet 10.1.1.5 netmask 255.255.255.0 up group lan group pub"

gonna reboot and test again

ok even with that change service netif restart still breaks networking

rtprio: ok i found something else. after the service netif restart, bridge0 lost 1 of its 4 members, the igb0 interface

what would be causing that?

lw: if service netif restart and service routing restart doesn't give you a working network, then that's a bug age needs fixing. locating it, however is a different matter……

meena: in principal, you are right, but... :-)

meena: well i've got it pretty narrowed down here

am i missing anything or do i already have a repro?

(weird network behavior after running 'service netif restart' is a freebsd bug or? termbin.com/sgm1b)

lw: i only know one person who's read the netif code and come out not wanting to rewrite all of rc

termbin.com/reux updated with the missing member line

i will get around to writing my dladm one day

In not sh, i hope

probably C, although i guess the chance of something like that ever being in the base system is around zero, so language doesn't really matter

(any language except sh, anyway)

should rewrite it in rust if you want it to be really good

lw: we have a C, C++ and Lua in base

surprised javascript hasn't made it into base yet somehow

devd, at least, is C++. there's probably more

C++ might work although for reason i prefer C for this kind of low-level stuff, it just feels more appropriate. that doesn't really make sense...

lua is just like JavaScript only if it was designed with care

so anyone know what's causing the networking bug in termbin.com/reux ?

that is a confusing looking config

the top comment explains what i'm trying to do. it's been working just fine for me before needing to service netif restart

i'm open to trying changes if it improves things though?

how would you do the network config instead?

lw

you forgot the ":" to ping :P

jbo

whew [main-default] [2024-01-13_07h57m36s] [committing] Queued: 825 Built: 825 Failed: 0 Skipped: 0 Ignored: 0 Fetched: 0 Tobuild: 0 Time: 14:14:27

ok, now what: le-fay.org/tmp/30d/uBIL0d.txt ... i think i remember running into this error before and it's caused by pkg doing the "running kernel" check in a really weird way

i'm running c8328f1a7b6e and __FreeBSD_version was bumped to 1500010 in a2da1bdb61bc2 so this cannot be right

ah right it's caused by building world WITHOUT_CLEAN: freebsd/pkg #2162

Title: determination of running FreeBSD kernel is bogus and misreported [pkg 1.20.4] · Issue #2162 · freebsd/pkg · GitHub

2162 – New port: scsh - The Scheme Shell bugs.freebsd.org/bugzilla/show_bug.cgi?id=2162

can anyone recommend the labwc compositor?

how do i view which packages i have installed on my system?

"pkg info"

Use "-x" option to specify a pattern

related manual page is "pkg-info(8)".

thanks

weird network behavior after running 'service netif restart' is a freebsd bug or can improve my config? termbin.com/reux

I am wondering how to get FreeBSD to dual boot with archlinux from archlinux /boot/EFI

if i start the `service seatd start` will this also be started on the next reboot automatically?

AumShivaya: I did that yesterday on a BIOS GPT device

I think you want to use the enable command

`service seatd enable`

to enable at boot

oh? how did you go about it?

dmr104: if 'service seatd start' works then yes, because freebsd won't allow you to 'start' services that aren't enabled - you have to use onestart for that

I installed freebsd first, then archlinux with grub with an alteration stub to the grub config mentionin the UUID of the bsd partition

dmr104: if start doesn't work then sysrc seatd_enable=YES

oh you used grub

hmm I am using systemd-boot

I find grub difficult at the best of times

but they are all difficult I suspect

it is powerfully amazing though

hmm ok, I may ditch systemd-boot and give grub a go

did you just point it to the BSD partition?

did not have to copy things over to the arch boot partition?

iam.tj/prototype/guides/boot is a good read

Title: PC: Illustrated Guide to GRUB and Linux Boot Process on BIOS and UEFI

or merge them, somehow?

I put a command like "freebsdboot /boot/loader" in the /etc/grub/grub.cfg Grub knows where to look on the partition

sorry /etc/default/grub not grub.cfg : that is something else

i got help from #archlinux, but you need to understand about disk partitions. this took me some time

an alternative is to get freebsd boot loader to recognise the archlinux, but i don't know how to do that

why does `man service` refer to start/stop/etc as commands for `service seatd "etc"` ? What ambiguity is this?

dmr104: what part of the manpage are you referring to?

"control (start/stop/etc.) or list system services"?

8

no, i mean, which part of the text of the manpage do you find ambiguous?

how would i know that "enable" is a command?

dmr104: rc(8), the section starting "Each script is expected to support at least the following arguments". those commands are provided by the script itself, not by service, so service doesn't know what commands are available

there should probably be an Xr from service(8) to rc(8) to explain that...

meena can you tell me what's confusing about my config plz?

I have a box at 12.1. Can I skip directly to 13.0, or do I have to freebsd-upgrade 12.2, 12.3 etc. ?

dmr104: hopefully this makes it more clear freebsd/freebsd-src #1057/files

Title: service(8): direct user to rc(8) for a list of valid commands by llfw · Pull Request #1057 · freebsd/freebsd-src · GitHub

lw: looks good.

p/

odd to see people with @suse.de email addresses in the git log