anyone have a jexec bash completion?

anyone know how to program bash completions? really really want completion for jexec and jls to complete jail names based on any configured in /etc/jail.conf.d/*.conf

whether they're running or not

i'd learn how myself, i'm not lazy, i just don't have it in me to learn yet another lang (bash) for 2 completions

and bonus, if you help me you can sell your jexec and jls completions in the pkg store (for free)

so everyone will get to use them when using jails tech

alepzi, It's a bash specific syntax and for bash "." and "source" are the same thing. So it does not matter. Both are equivalent. The ". file" syntax is standard /bin/sh and so bash gets that from there. The "source file" syntax originates with csh and bash pulls it in from csh.

how do I identify where upstream is for contrib/nvi?

so someone on mastodon has come across a drive (memdisk specifically) exhibiting both slices and partitions

link?

how does one find the current pkg build state of a specific package?

clairvoyance

how can i get same package list installed on new station as on old one?

etwas: on the original machine, you can run "pkg query -e "%a = 0" %o > pkglist.txt" and then copy that file to the new machine, and run "xargs pkg install -y < pkglist.txt" (please review the file before running the second command to make sure it is sane)

realdeimos: do you have a machine with IPv6 routed to it? If so, you can use pkg-status.freebsd.org/builds?type=package&all=1 to find the specific builder jail for the version-arch tuple that you're running, and click the little bomb icon next to the build you wanna look at, that'll take you directly to the builder.

The builders are IPv6-only, because typically all they do is send built packages to the package CDN once it's done building.

sfox: that's not really helpful at all.

LxGHTNxNG: that sounds like a disk with an old-style BSD label?

i'm currently having a hard time setting up a second, separated LAN. i set up isc-dhcpd and coredns and those components work fine. my client devices get an ip in the right range and the dns info. the server itself can reach the internet fine (it sits behind one of those enduser "router" boxes), but clients in the separated lan can't reach the internet.

my understanding was that gateway_enable and a default route were all that's needed for that to work, but i can't find any specific docs on what gateway_enable does in detail.

doing dns queries from clients in the separated lan works fine, i essentially just need the one line to tell the server to actually route client traffic to the upstream network.

phryk: gateway_enable enables IP forwarding, which means if the kernel receives a packet on one subnet with a destination address in another subnet, it will forward the packet. to start with, check the clients have a correct default route, and see what traceroute says

yeah, the default gateway is delivered via dhcp. it's the ip my server has on the interface on the separated lan. traceroute gets to exactly that host and then sits there spitting out asterisks

this is IPv4 traffic, right? there's a separate gateway setting for IPv6 traffic

yes. ipv4 only. upstream has only v4 so i saw no sense setting up v6.

does the router have a firewall running (ipfw/pf/ipf)?

my understanding of "enabling ip forwarding" was that if the server knows a route to a network (i.e. it can reach it itself) it will be able to forward client packets.

lw: usually yes, but i deactivated pf for testing.

yes, that's how it works

lw: okay that's weird because the server can clearly reach the upstream lan and the internet, but the clients can't.

is there maybe some weirdness because i set both gateway_enable and defaultrouter in my rc.conf?

phryk: run tcpdump on the router's outgoing interface (the one facing the internet) and see whether the packets are being sent but no the replies, or if nothing is going out at all

no, it's fine to have both of those set

k sec, lemme look up how to do src filtering with tcpdump again^^

phryk: 'src host 1.1.1.1 or dst host 1.1.1.1' should work but i'd recommend trying to test this without filtering if possible, just in case something odd is happening

(not always possible if there's a lot of traffic though...)

ah yes, not wise to do this via ssh… thank you, feedback loop.

if it's just ssh you can try port not 22

(assuming your test traffic is not ssh :-)

na, just ping for now

okay, i can see "IP $clientip > $pingdestination: ICMP echo request…"

and no reply?

responses would be "$pingdestination < $clientip", correct?

no, just >

no, the response would be 'IP destination > client: ICMP echo reply'

ah, okay. but no, only echo request

same behavior for the upstream gateway (192.168.1.1) and hosts behind the nat of that gateway (wikipedia.org)

so it seems like the router is forwarding the traffic fine and it's being dropped later, perhaps at your "end user router"... if that's doing NAT, does it know to NAT the specific address space in the new subnet?

unless it just answers everything, no. but access to that box is lost, that's the reason i'm building a second lan.

so i guess my only option is introducing a second nat on my server?

if you need NAT and you can't configure the edge router to do the required NAT - probably, yes. are you intending to replace that router with this new freebsd router though? in that case you'd presumably need to do this anyway...

when i switch to a good isp, yes. i'm working based on my old setup where this was the edge router/fw anyhow. :P

but current uplink is from my roomie and they lost the pw, so in the interim there wasn't much else i could do except try and run a dictionary/bruteforce attack against the routers web ui thing and that seemed inefficient^^

Hi, I did a pkg upgrade on my FreeBSD 14 and all the graphic applications stopped working

Firefox, etc. Now XFCE is broken

is there a way to restart it without rebooting?.

martinrame: log out, restart display manager, log in again. could work, should work, might not work. :P

phryk: let's try that

lw: reactivated my old nat rule and it immediately worked (written from formerly problematic client) :P

phryk: another option would be to just bridge the two networks, which may or may not be easier depending on what your goal is

phryk: mmm, the login manager shows "faled to run command" (or something similar). I had to ctrl+alt+f2, then login without seeing anything, and run startx, now I'm on XFCE again, but I'm afraid I'll have the same issue when I reboot.

btw, thanks to tmux I was able to re-attach this session ;)

martinrame: if you're on a tty anyhow you can try unloading and reloading the graphics driver module.

tho a reboot would also make it load the new driver^^

phryk: but I am on a graphic session right now. The issue is with the login manager

Mmm: pkg: libxine has a missing dependency: libva-glx

yeah, no idea. i'm still sad slim broke for me and i had to set up lightdm which still confuses the hell out of me^^

phryk, how did slim break for you? For a week or so a month or two ago, it displayed a blank screen for me, but that has been fixed about a month ago.

Soni: your netinet6 issue was fixed in main, you can cherry-pick this commit: freebsd/freebsd-src 042fb58 - you may also need 0478a0356272ab3907af9d97711f0bb8b14bf38b 122dd78c143e3d138b5d503a30da4c326f41002a

Title: sys/netinet6/in6_pcb.c: fix compile without INET · freebsd/freebsd-src@042fb58 · GitHub

Soni: there are also two unmerged PRs still going through review, freebsd/freebsd-src #1159 and freebsd/freebsd-src #1157 but you fix those by disabling the affected modules (if_bridge and pf) if you don't need them

Title: sys/net/if_bridge: support non-INET kernels by llfw · Pull Request #1159 · freebsd/freebsd-src · GitHub

(the pf fix should be safe, if_bridge i am not sure about and would like more review on)

lw: ty

debdrup: thank you. I was poking around earlier, but still can't find the specific package. It appears I have to click into each build and then search for the package

the port is current as of 06 april, just waiting on package to be built

and i'm not using quarterly packages, using current/daily whatever it's called for pkg repo

latest pkg repo is what I'm saying

so i got a separated lan NATing to the upstream lan and having working dhcp. i also got dns with coredns set up, and it works fine on the client machine in the separated lan. but vms (using bhyve-vm) seemingly can't talk to the dns server – even tho they *can* ping it and get dhcp leases from it just fine.

phryk: are you using a bridge(4) for the VMs?

lw: no idea, whatever "vm switch" creates.

ah, i personally dislike vm switch precisely because you have no idea what it's really doing :-)

phryk: can you paste the output of ifconfig -a?

it's not *called* bridge*, but i'm not sure how much of an indication that actually is. since i just added the lan interface to that with (vm switch add) it might be a bridge…

sure, sec

i'm fairly sure 'vm switch' creates a bridge, but i'm not sure how it configures it

seems like tap

:P

tap will be the VM interfaces, 'vm-public' is your bridge

phyber: you need to move the IP address of the host from re0 to vm-public, and adjust firewall rules as appropriate - unfortunately i'm not really sure how you do this with vm-bhyve's bridges

mhh, i thought it was either one bridge or separate tap interfaces per host.

that may not be the cause of this particular problem (although it's possible) but it is wrong to configure IPs on bridge member interfaces

sec, gonna adjust rc.conf

phryk: no, it's both - the tap interfaces provide the packets from the VMs, but without a bridge, they won't go anywhere. the alternative is to not use a bridge and instead route the traffic (i.e., assign a separate /30 or something to each tap interface)

tykling what is required for my freebsd 14 jail to have access to pool/backups inside my pzbackup jail? im using this in my jail.conf exec.created += "zfs jail pzbackup pool/backups"; allow.mount.zfs; enforce_statfs; but i keep getting cannot set permissions as root inside jail when i want to set zfs allow on the pool for my user .. any idea

lw: well, that did not work…

for one, '-' in interface names is something rc.conf really doesn't like, even with escaping. but even when i manually set it up so that re0 was just up without lease and did dhclient vm-public it did not change the reachability of dns in the vms

oh no

phryk: like i say i'm not really sure what the right way to configure this is with vm-bhyve's default bridge setup :-/

i just create the bridge by hand and tell vm-bhyve to use it

Anyone else experiences /bin/sh occasionally "forgetting" that it's in vi mode, and entering a really strange mode where you e.g. cannot edit/delete the very last character, as the cursor won't move past it?

Dooshki: no, but does ^L help?

Interestingly enough, issuing "set -o vi" fixes it and makes it behave good again

/bin/sh has fairly primitive terminal handling and its line editing gets confused if there's any output before the prompt

huh… uh, wat. i added the log directive to the coredns conf, saw the queries arriving but vm didn't get a response. start coredns again and now it works, at least for the debian vm? o_O

lemme test the dinwoze one

lw: No, ^L does literally nothing

phryk: you should fix this anyway, having IP address on bridge member interface is wrong. it will cause uncommon but hard to diagnose problems

(it prints the character)

(and you're able to run it as a command)

Dooshki: ok i see, ^L doesn't work on vi mode. not sure then

maybe there's a vi-mode command to do the same thing

lw: Actually, it does work in vi-mode, it just doesn't work in the bugged in-between mode

when I type set -o, it lists "vi mode" as enabled

really? i just tested it here and after set -o vi (and nothing else) it didn't respond to ^L

lw: You have to exit into command mode first

ah

pressing <esc>

this is different to how zsh works, i never used vi mode in sh

I guess this is another idea for a weekend project, seeing if I can debug this

probably not this weekend though

oh this is annoying, because FreeBSD-utilities has its own pkg ucl, it doesn't automatically pick up library dependencies on things it needs

so i moved libpjdlog to another package and now: /sbin/decryptcore: libpjdlog.so.0 => not found (0)

i thought pkg was supposed to handle this automatically even if you don't explicitly list the dependencies though...

okay, i booted the windows vm and now it works on neither the windows vm, the debian vm *or* the host… i think coredns is just plain broken. :/

ah i see the problem: it installed the lib32 package instead

tykling was missing the allow.mount, allow.mount.zfs wasnt enough

lw: lol, it works when i deliver the upstream dns via dhcp… but not if i deliver my own dns in there and set the upstream dns manually in the windows vm. go figure…

finding some method behind the madness: coredns either works for the host or the vms, and apparently whoever first queries the server gets dibs…

mhh, it looks like it works fine IFF i specify an interface to bind to…

phryk: did you fix the IP address being on the member interface? because this is exactly the kind of issue that can cause

i have 1 nvme ssd installed, so i know nda0 correlates to nvme0. how can i systematically know that, i.e., is there a command that shows that relationship?

markmcb: as far as i know, no. i'm not sure you can hardwire this with device.hints either. if you're trying to script something, i suggest using GPT labels instead of relying on device names

maybe camcontrol devlist

nda(4) suggests camcontrol devlist but i don't see this info there

<WDC WDS100T2B0C-00PXH0 211070WD> at scbus8 target 0 lun 1 (pass1,nda0)

you might have to use inxi or parse dmidecode

anyone know how to program bash completions? really really want completion for jexec and jls to complete jail names based on any configured in /etc/jail.conf.d/*.conf, without them having to be started and running. can't bring myself to learn enough bash-completion to do it but i'll bet if you do it other ppl would use it (pkg install jail-bash-completion?) not just me

alepzi: have you tried #bash ?

tried yesterday didn't get anywhere

not too many freebsd users didn't seem like

Until 14 the default root shell was csh so knowing that it might explain why not so many bash completion enthusiasts present here.

rwp: as if anyone but 80 year old greybeards actually used csh :-d

I have used tcsh a lot:P

tsoome: what colour is your beard?

em, lets talk about something else..... :D

At a previous employer the entire design lab used tcsh almost exclusively. It was swimming upstream to use anything other than tcsh.

okok, not that grey:)

alepzi: maybe ask on questions@ in case someone has already done this

alepzi: if you do find the answer, you should submit a port for it

I am still tripping over the use of sh for the root shell because my fingers are used to bash key bindings and not all are supported. I need to spend some time with the new sh key binding docs and figure out how to do one mandatory thing that I use all of the time before I can use it.

rwp: do you ever actually log in as root? i normally use sudo -s which gives me my zsh

In sh command line how does one recall the last argument of the previous command? In ksh it is ESC-_ and in bash it is either that or ESC-. to past in the last argument of the last command. I use that all of the time.

Using "sudo -s" is incompatible with "sudo -i" and "sudo -s" leaves HOME set to my home, which inevitably drops root owned files in my $HOME which then I need root to clean up afterward. So... No. I never use sudo -s. I only ever use sudo -i or just sudo command.

sudo echo $SHELL tells me bash

Having root owned files in my home has snagged me regularly enough that I have a personal cron "test ! -d $HOME || find $HOME -user root -ls" to look for accidental files and report them to me.

alepzi: 'sudo echo $SHELL' will print exactly the same thing as 'echo $SHELL' because the $ is interpreted by your interactive shell

yikes

what's the right way to echo $SHELL as root then?

rwp: interesting, i haven't run into that issue. 'sudo -s' sets $HOME to /root for me

alepzi: sudo sh -c 'echo $SHELL'

You might try "sudo printenv SHELL" to avoid dollar expansion or quoting at all.

[25!] ilythia /src/freebsd/src/pkgbase-share-examples [lf/pkgbase-share-examples]% sudo sh -c 'echo $SHELL'

/bin/sh

[26!] ilythia /src/freebsd/src/pkgbase-share-examples [lf/pkgbase-share-examples]% sudo echo $SHELL

/usr/local/bin/zsh

csh!

sudo -s definitely sets HOME=/usr/home/rwp here.

For those lurking along try: sudo -s env

that's odd

% sudo -s env|grep HOME

HOME=/root

maybe we have different sudoers(5)? but mine is basically just the default with some extra rules

Hmm... I have "Defaults env_keep += HOME" in the sudoers file here. Where is the origin of that? Looking...

The origin of that for me seems to be lost in the sands of time. But probably it was a default setting at one time. I also have "Defaults env_keep += MAIL" too.

I am deleting the "Defaults env_keep += HOME" line and will see if the source of my complaint was self-inflicted all along.

i don't have either of those but i suspect that's the cause of your problem

I think you are correct. (But I am going to keep my crontab looking for errant root owned files anyway.) :-)

But anyway... I do often log in as root. And for example the last day I have been working trying to get IPMI SOL (serial-over-lan) working for (ahem, sorry) the Linux kernel. And decided that I just can't make it work with Linux. I think they have removed the ability.

The Supermicro BIOS works over IPMI SOL perfectly. GRUB works over the IPMI SOL perfectly. But Linux insists on using the graphical console rather than the hardware console and so far I have failed to make Linux work on the IPMI SOL.

The task is to get it going with Rocky 9 but I am going to take a diversion and fire up a FreeBSD system there to see how that works out over the IPMI SOL.

I feel certain that FreeBSD won't have this problem. But I will find out in an hour. Lunchtime here first though.

i would be quite surprised if freebsd handled ipmi serial better than linux, but you never know

although if it's a case of configuring it to use serial, yeah, freebsd is probably easier. always a hassle doing that on linux

anyone try nushell on freebsd?

Hi, I am considering buying a laser printer, I have a Samsung almost in new condition, but the driver is a pain. If you buy a Brother printer with PCL6 emulation, does anyone know if it would be able to work with LPR -no CUPS-

uskerine: yes it can, but it depends on the application you're using being able to generate PCL output

by LPR I mean LPD

I will only ever buy a network connected printer ever again. I am done with USB connected printers that require closed source drivers.

you can buy a tplink printer server

and attach it to any printer, then it is ethernet

where I have had issues is with the drivers when you do not want to use cups

HL-L2370DN I am checking this one, but I fear that the PCL6 support is a fake and then it is in reality done by the software driver

I am not recommending any specific printer because I have been very frustrated with most of them.

Rage against the Machine was never really specific about which type of machine they were furious with but I think it was probably a printer.

Agreed, same here

except for my Star EPSON compatible, that things prints everywhere, plus it is cool seeing the continuous paper

The Epson has tractor feed paper? That's so retro!

I looked at the HL-L2370DN that you mentioned, I see it says it has Linux drivers. I tried to download the driver to peek at it and the web page them breaks for me and I can't download it.

I suggest downloading it and unpacking the rpm/deb and inspecting it to determine exactly how they are shipping their driver for the printer.

If it is a binary only driver then I would run away from the printer.

the best thing would be a PS printer, that requires no driver

but they are no longer common

I think the most expensive ones still have it, but even in that you can not be sure a punk put that in the driver and not in the printer

So one of my clients has a Brother USB printer that he had previously been using on his Ubuntu system. Well it turns out that the Brother drivers were 32-bit binary drivers for Ubuntu 16.04 *ONLY* and nothing else would work. PITA! Note Ubuntu dropped 32-bit support entirely.

I ended up creating a VM running 32-bit Ubuntu 16.04 VM as a printer spooler. It's an appliance for my client now.

For my samsung I was almost sure that the driver was almost there, but somebody does something 10 years ago and the thing gets packaged and repackaged and it is no longer simple as ti shall be -so you can use the filter in LPR and that is-

that is a solution, but I am sure you do not like it that much

This is NOT a recommendation but I own an HP Color LaserJet MFP M283fdw network printer which is supported by HPLIP and works for me using the network interface. It's a native PostScript printer.

It also scans and I have it configured to scan and email the PDF scan to my other system on my LAN which then has a procmail rule to unpack the attachment and drop it into my NFS scan directory.

I do not want a color one, they are expensive. I just want to print my .PS diagrams and some PDFs

.ch

also new ones say they support PDF, does that mean that you can drop a PDF without any driver/filter from LPD?

I have also owned a series of HP InkJet printers over the years. I will never buy another one. The last two HP InkJet printers I have owned were endless trouble from start to finish. Never again.

uskerine, just wait for the next one to show up on daydeal

they come and go every couple of weeks/months

inkjet printers never last, the best printer is the one you do not have, but I miss printing some diagrams from time to time

lw ping

jbo: urrrrrrgh nooooow whaaaat

jbo: hang on i have to fetch more alcohol, ack in 10 mins

lw, regarding 276996, does that even still make sense given that you dropped maintainership of your other ports? we don't really want to add ports that we "know" will be orphaned in the forseeable future.

hey, I'm using auto-admin which is looking for meta.txt in pkg.freebsd.org/FreeBSD:14:amd64/quarterly but it's not there. Now... I checked a couple of other directories in there and they have a meta.txt including pkg.freebsd.org/FreeBSD:13:amd64/quarterly. So... is there an update that make it so meta.txz is no longer used?

Title: Index of /FreeBSD:14:amd64/quarterly/

looking for meta.txz, sorry, I'm working with text files for work, typo there.

jbo: ok to you, if you don't want to commit it i will just maintain it locally

s/ok to/up to

lw, the question is whether you have any intention of abandoning it as of now.

jbo: my frustration was mostly aimed at python packages, i think i'm okay to maintain to a single package with no dependencies, but honestly, no guarantees

my opinion that ports is completely fucked has not changed

is it just me or does apropos not parse the document description of any man(7) pages?

is anyone willing to check "apropos -s 1 dc" and/or "apropos gnupg" for me?

concussious: looks fine here: le-fay.org/tmp/30d/4WWdYg.txt - although i agree in general trying to find manpages is not easy

lw: thank you so much. that's what it's doing on my fx14p6 which is completely wrong

the description for dc is dc - arbitrary-precision decimal reverse-Polish notation calculator

ah you mean the output from apropos

bsd style man pages search really well

yeah, that seems wrong: dc(1) - dc

but linux style man pages are broken in our implementation

gpg is worse, it doesn't come up with anything at all

I don't have gpg installed, so I can't tell

do i mail the mandoc list? if this can get fixed, manual discoverability will be a solved problem

concussious, Did you find your question about the upstream of contrib/nvi?

idk where the right place to report that is off hand

i guess man/apropos are from some upstream

rwp: yes and the maintainer accepted my patch immediately

man and apropos are from mandoc.bsd.lv

Sven? Awesome!

well, man is actually our own shell script

but the rest of it is comming from mandoc.bsd.lv, those guys are using cvs and mailing lists, i don't yet understand how to interact with them politely the way they want

i got some params that are in all of my /etc/jail.conf.d/*.conf files. can i move those to /etc/jail.conf and keep unique params in the individual jail configs?

rwp: no vi upstream is lichray/nvi2

Ah... Yet another fork in the road.

don't remember what it was, probably something super trivial

lichray/nvi2 #121 ah yeah

Title: fix last left-over K&R function declaration by igalic · Pull Request #121 · lichray/nvi2 · GitHub

alepzi: probably, try it and see

lw, I'll consult and report back - I prefer to ask $mentors about this one

(nothing personal, hopefully obviously)

jbo: this would be easy to fix if you just gave me a ports commit bit

although tbh... not sure if i would take it

people repeatingly asking for commit bits are AFAIK not the ones necessarily getting them :D

sounds good!

can i have a commit bit pls

are they trying to backdoor freebsd?

oh you won't let me participate in this nonsense? too bad

alepzi: absolutely

that's what i think

lw, too much or too little alcohol rn?

jbo: too much ports. it annoys me.

commit bit is a responsibility noone should want, but they do for good of the project

my life has become notably more miserable since I got one. and I haven't done ports work the past few weeks :s #feelsBad

alepzi: that's kind of meaningless though because anyone with a commit bit should want to contribute to freebsd

otherwise like, why would they have it

duty

lw, I can't wait for space age, seriously

Volunteer work is the same everywhere. There is no minimum. There is no maximum. It is up to everyone to manage their own burnout.

haha, duty

-__-

never heard of service beyond yourself lw?

lw, given that you're already in rage mode - would you mind telling me why a python thing is no longer running/working? I tend to stay away from python so...

alepzi: yes, i've heard of it, i think it's nonsense. used variously by neocons, neolibs and marxists to argue why people should conform to their political ideas "because it's for the good of all"

lw, starting cad/freecad shows the window but this error message: pastebin.com/Y5yVsuje

Title: 22:29:38 pyException: Traceback (most recent call last): File "<string>", li - Pastebin.com

jbo: uuh, really not sure. looks like an issue with a C++ module. have you tried recompiling all packages?

lw, yeah, this cad/freecad comes from my poudriere repo

jbo: did this happen after you switched from 3.9 to 3.11?

lw i'm sorry ports hurt you in your nono spot. but you can overcome i believe in you!!

alepzi: why would i want to overcome it? it's much easier to just maintain my own local ports tree

alepzi: i am not the only person who has issues with how freebsd ports, works fwiw. although i won't name names.

s/ports,/ports works,

somebody forgot a \

jbo: you might try pkg upgrade -f just incase somehint didn't get upgraded properly

well ya i think it's a problem pkgs doesn't keep the most 2nd most recent version of a package around. that's a big prob

you're talking a shitload of storage

it's only twice the current amount

it would fix a big hole in the native freebsd pkgs system. that a new pkg can have a bug the old 1 didn't and there's no seameless way to rollback

"only"

pkg install nginx@prior, fixed

2x is an easy multiplier in tech don't sass me

feel free to ask clusteradm how much space a single set takes up, and how doable 2x that is distributed across all the mirrors

they ever come on irc?

other than that, ABI issues.

not public spaces that I've seen, you'll need to email

actually, bapt might be ablr to grab a quote and he's around

i got into an argument with postmaster@ the other month so now i can never email any freebsd.org admins about this stuff

bapt ^

well, not around right now. probably late in france

france is cool

alepzi: bapt isn't on this channel

i think he's just on efnet

lw, what did you do

jbo: ?

so we have ca 15000 ports, 7 architectures, 3 releases; I'm gonna guess we're talking about 4 TB minimum?

lw, w.r.t postmaster@ argument

jbo: oh, i got annoyed about how they do mail forwarding re: spf and sent a rather rude response

jbo: which was totally my fault

meena: twice as many ports

lw, ease off mate :D

ease of? I don't english very well

jbo: this was like two months ago so the off is eased

meena: that's a few HDs to throw at a permanent and total solution to the "omfg new pkg has bug in it but last was fiiiine"

a few HDs x machine x each mirror

build it and sponsorship will come

no datacenter already giving freebsd the homie prices will balk at a few more HDs

we can do it!!

no one is going to sponsor "pkg.freebsd.org but it's one release behind"

anyone who cares about that is bullding their own packages

I have an infant on my chest so I'm just gonna kinda step away and stare at him instead, better for my blood pressure

ya feels like @prior would be a nice middle ground

o/

[]

alepzi, here's an idea: you could sponsor it.

only if lw gets commit bit

i don't even commit to ports, how would that help

chaos engineering

i think the important thing to understand about pkg.freebsd.org is it's terrible and no one uses it

anyone who is seriously using freebsd builds their own packages with poudriere

I wouldn't call it terrible. It's very difficult to have a "sane default" setup.

pkg.freebsd.org is just a trap for new users so we can pretend we have useful binary packages

i don't think it's terrible it's pretty good tbh

just has this 1 hole i found

alepzi: let's say i was off by a bit, and we're talking more like 12-20 ; to duplicate that means you need to have enough disks in each machine (and enough spares) serving pkg.f.o, and you need to pay for the bandwidth of replicating it. that's a lot of money

jbo: sane defaults is arguable, but waiting 3 weeks to get a security update...

jbo: like the latest issues with armv7/arm64 on the mailing list, fricking hilarious

meena: :(

yeah, bulk builds are a btch

imo easy bin install is table stakes for an OS. call it marketing budget if nothing else

you're right that serious stuff should build their own pkgs

alepzi: if we're talking about onboarding new users, i agree. pkg is better than what we had but it's still bad

the first thing about hardware you gotta consider is that it can fail, so if you want to double your resources, you also need to consider your spares

but this will never be fixed as long as ports czars believe everything is great

lw, I have yet to encounter a developer (or ports committer) who seriously things that everything is great.

this isn't just a pkg issue, the entire ports system is broken

I was referring to ports

okay, another night, another zero lines of code written thanks to IRC

biggest 1-2 ways ports needs to improve?

meena: you can do it!!

alepzi: 1. rearchitect the 2. whole thing

alepzi: i can do sleep, yes

lw funny but concrete shit?

debian solved binaries packages in 1996

freebsd will never solve it because we have ports and it's "good enough"

still not telling me specifics

alepzi: fix whatever needs fixing that makes our ports collection rebuild every time curl does a minor or patch release

yeah honestly i don't care about convincing you about my views on ports

like a hook to autorebuild from source when a new version is tagged?

that would be building hell

lw ya you're just a burnout that wants to shit on freebsd instead of talking about how to improve it that sucks but i want the best for you you can do it!!

what we need is a binary repository

with a release history?

that's what debian did and largely what redhat do and it works great

what do you mean by binary repo?

alepzi: shit on freebsd? sorry, how many patches have i contributed to freebsd this month and how many did you contribute?

big fat 0 my man, wonder why

okay guys, cool it off.

i am happy to talk shit about ports because it's shit, but i also work to improve things

(or of - still not good at english)

what month is it even

November

I think i got 0 the last two months

jbo: good grief, which year?!

meena, 1997!

so, everything is actually quite good :)

pretty sure we still had war in Ex-Ju in 1997

yeah I guess that was not the best of statements I ever made.

jbo: is 4.4BSD out yet or are we still on 4.3BSD?

alepzi: consider this, you can't make something better until you recognise why it's bad

lw that's why i was asking you specifically the 1-2 biggest ways it could improve

you're blinded by hate, burn in it

[00:00:10] [01] [00:00:00] Building devel/cmake-core | cmake-core-3.28.3

oh god...

"1-2" ways doesn't make sense, the entire thing is broken

ports works basically the same way it did when i started using it in 1996

'biggest'

alepzi, many problems/issues arise from design/architecture decisions.

it's not like we can incrementrally fix this

ya i do that all day

it's a redo from start problem

it's like saying "tell me the two biggest ways of transforming this bicycle into a skyscraper"

lw: i think it might be worth looking at what OpenBSD did differently since… 1996 and figure out if we can replicate that

because they don't seem to have the problems we have

yes, i UNDERSTAND, and yet you're still not giving specifics. what do you aim to SOLVE with a complete reengineering other than hand waving?

rearchitecting

meena: didn't openbsd not even provide ports security updates until really recently?

(i don't use openbsd but i heard this)

lw: i have no idea; I've only started trying to add a port to OpenBSD this year, and it seems… fine. mostly

(i don't use OpenBSD either, but, i felt compelled to do it for reasons)

meena, I remember when I added my first port to FreeBSD things seemed... fine. mostly. too :p

alepzi: do you want me to sit here on IRC and describe a complete rearchitecing of the ports framework? ... feel free to post to ports@ or something if you and care and i will respond there

from my limited understanding: OpenBSD doesn't try to do everything in Makefiles; when building Python stuff, they actually use python code

this is not a trivial problem i an explicate in 512 bytes

s/i an/i can/

ya more blowoff. not surprised

dazzle me with your engineering fu

moved allow.raw_sockets;, devfs_ruleset = 5; and vnet; into /etc/jail.conf from each jail's config in /etc/jail.conf.d/*.conf, and starting jail fails: cannot start jail "testjail": epair42a, 1, ifconfig: siocsifvnet: file exists, jail: tstjail: /sbin/ifconfig epair42b vnet testjail: failed

i'll try putting vnet back in individual jail configs

ok that worked. so allow.raw_sockets; and devfs_ruleset = 5; can both be in /etc/jail.conf, but vnet; has to be in the individual jail config

looks like vars can't be in /etc/jail.conf if they're used in /etc/jail.conf.d/*.conf

exec.clean; can be in /etc/jail.conf

appreciate the hard work. this OS is second to none

exec.consolelog can be in /etc/jail.conf and it can even have the ${name} var in it!

ya freebsd is the best OS ever

:)

full customizable

yum

exec.prestart doesn't seem like it can be in /etc/jail.conf and global for all jails :/

alright got my jail setup all tight. ty everyone! and even lw

man zfsconcepts doesn't have a section on what a Volume is, or what a File system is, or a Dataset