meena: if you wanted to build a source package today, for a system using quarterly packages, would you built from 2023Q4 or 2024Q1?

the correct answer is 2023Q4, because 2024Q1 binary packages aren't ready yet. but how is an end-user meant to know that?

I… would not build packages for quarterly… i am using latest everywhere,

sure, but most people (who use binary packages) are using quarterly

that sounds fake

like, most dogs who go outside do that for eating slugs

i will stand by this fact, since quarterly is the default

(that's probably just my dog)

It's the tyranny of the default. I mostly use pkg, which defaulted to quarterly, having no reason to change it I have not, I am still on quarterly. It's been fine.

i wish ports could build subpackages. like pulseaudio and pulseaudio-jack

this would solve a *lot* of cases of building from source

unixwitch: oh! ooooh!!! subpackages *just* landed.

sub-packages?

meena: oh really, i saw something about this but i thought it was for pkgbase for some reason

if this is true it might be enough to get me to switch back to packages

Title: ⚙ D40549 Subpackages!

hm, www/qutebrowser says it's like vi, so i tried :q to close a tab and it quite the entire program

:bd

the previous review reviews.freebsd.org/D16457 has examples of how to use it

Title: ⚙ D16457 Subpackages. (deal with it.)

"Subpackages. (deal with it.)" darned greybeards, eh

meena: this looks great, i hope everything gets updated to use it this week!

it's already friday here, and that means two things: I need to bring the dog out for a piss before going to bed, and then try to sleep

So do subpackages basically mean that a package can be built as a single unit but broken out piecemeal during install/deinstall?

Oh

_xor: it means you can build a single piece of software and create multiple packages from the files it installs

so instead of using on/off OPTIONS, you just build mypackage, mypackage-foo and mypackage-bar

most importantly, you can break out docs/man pages

so if you're installing tiny … things, you can leave those out

i actually don't care about that but i can see it would be useful for embedded systems

everyone benefits!

That would be very nice, yes. I currently build without DOCS generally (since I install a lot of packages on my servers), and just use the web for docs when needed.

although, i hope there's a way to automatically install man packages?

Every once in a while I think about rebuilding the repo with DOCS set, which would probably take up a 1-2 days.

think headers, docs, man pages, etc… you can build really tiny jails. I can now see why this is pushed to conclusion by pizzamig, they work on pot

maybe i'll make a patch for pulseaudio to do this, but otoh, i feel like whoever maintains that is probably already doing it?

The smallest base-runnable jail I built was around 18mb, I think.

unixwitch: this is desktop@ and they are buuuuuuusy… so any help is appreciated, I reckon

_xor: *nice*

i feared i might need to sub to desktop@, my poor inbox

i think dfr's are even tinier, somehow. I don't know how

I don't use it though, because so many packages rely on more than just that 18mb from PkgBase.

With sub-packages though, I might be able to do so. We'll see.

unixwitch: i am subbed to wireless, for some reason, and I don't even run FreeBSD on my laptop anymore (because I bricked it)

meena: laptop under warranty?

It's going to really help with packages like www/nginx though. Break out all of those modules and their individual dependency trees.

kenrap: not by a decade, also it's not a real brick, only a metaphorical one, because it's heavy work making it not-a-brick without complete reinstallation

AH

_xor: yeah. i hope postfix can benefit from this too although i suspect postfix builds this into its main executables

oh, yeah, i remember mwl complaining about this

Heh, I remember when I was younger I got the components to build a new system. It was an older AMD CPU, one of the ones that ran *really* hot. I was really excited to get a new system.

Put the thing together and I wanted to check that it would actually power on without issues, and without thinking, I turned it on...without the CPU cooler installed.

Apparently the new ryzen 7000 cpus run very hot even though they're very energy efficent

It was dead quite, the fan spun up for a second, and then I watched a tiny puff of smoke rise off of the CPU, and then the system died.

It was dead quite for a solid 5 seconds after that, with me just staring at the (now dead) CPU lol.

_xor: i did that once too. *once*

i can't get my Ryzen 5800X3D to go over 60C no matter how many poudriere builds i run, but i think this is because the X3D parts have a lower boost clock

lol I still laugh when I think about that.

unixwitch: join a google meet

Or a zoom meeting

meena: never. but i might need to find out if Teams work on FreeBSD at some point...

brrrrr

how is *that* better?

(or else figure out how to pass the mic through to my bhyve windows vm)

meena: it's not but it's what we (and all our clients) use :-/

The X3D gives it more cache, so maybe the cpu doesn't have to work as hard?

kenrap: i think the issue is the extra cache is on top of the CPU die (that's the 3D part) and they don't want to fry it by dissipating too much heat through it

Ah, haha

ok, most difficult problem of today, how the heck do i tell alacritty to open links in qutebrowser instead of firefox. this is some xdg crap, isn't it

ah xdg-settings

why can't we just use .mailcap for this smh

freebsd survey is about to close, if anyone cares (and didn't already do it) surveymonkey.com/r/freebsdsurvey24

Title: FreeBSD 2024 Community Survey

mailcap is for mime things

I would like to see the results of the survey

(already done it a while back)

kenrap: I reckon you will read about it in the next status report, at the very latest

rwp: xdg-open is also for mime-types, it just gets abused for url schemes by using fake mime-types

rwp: ok well maybe not mailcap, but there should be a plain text file for this

Probably not the next one, which should come out sometime this month, but the 1Q2024 one, in April.

i am confused that there's been a resurgance of non-"desktop" applications in X/Wayland, but everyone still relies on this xdg crap. even Sway needs it!

I replaced xdg-open by my own script, because I didn't want to run that code

hmm, that sounds like a good idea for a port

well, the settings are instde the script, because the handlers are python functions 0x0.st/H6vN.bin

My onboard sound sucks. I want to get a proper dedicated Sound Blaster card. Sadly I have only 2 PCI-e slots. One is used by wifi+bluetooth card. THe other, obscured by graphics card. Sadly, I do not think one can buy a PCI sound card nowdays new :/

AumShivaya: get a USB audio interface. better physical interface and, from what i've seen, better support in FreeBSD as well (unless you go with a high end RME card)

*sigh* all these dongles

and cards

what is an RME?

thanks for the advice

RME is a manufacturer of (among other things) PCIe audio I/O cards designed for professional musicians / producers

but who makes the most compatable usb sound cards for F'BSD?

sounds expensive

yes, they are quite expensive: thomann.de/gb/rme_hdspe_raydat.htm

Title: RME HDSPe RayDAT – Thomann UK

USB interfaces are 98% as good for most purposes and much cheaper

consumer-level PCIe sound cards are basically a dead market since everything has hda on board now, but there's a big market for lower-end USB interfaces for hobbyist musicians, and those tend to work well

and being USB isn't really an issue since USB has more than enough bandwidth for audio

ok

I just find that my onboard sound is terrible

optical doesn't mix audio streams

3.5 mm jack crackles loudly like a few times a day

in general, or on freebsd specifically? i thought modern onboard hda was usually ok

bluetooth is a pita

These issues above exist on BOTH Linux and FreeBSD

I do not know though, if the issue is my onboard sound conclusively, or my Creative Stage soundbar

i've always had back luck with 3.5mm jacks. all my headphones are 6mm, except for the earphones i use with my ipad

unixwitch, got the approval for the first two prs

jbo: i saw! exciting times

jbo: sorry i forgot to add it to subdirs :-(

unixwitch, no problem. it's 100% my fault that I didn't check. but that is what reviews are there for :)

there are git hooks that check for this when you create a new port commit but as I painfully learned those hooks don't get invoked when you import a commit :p

perhaps it is time for a new computer

you'd think make could create a list of subdirs automatically tbh...

but again - 100% not your fault.

jbo: do you happen to have a link to those? i always commit ports to a branch locally so that might be handy when i submit patches

unixwitch, they are in the ports repo - you already have them.

look in the .hooks dir

git config core.hooksPath .hooks or something like that

and you're good to go :)

unixwitch, also, use of DISTVERSION is prefered over PORTVERSION wherever possible. Not sure whether you checked the reivew but I modified that.

ah interesting thanks (re: hooks)

unixwitch, also, much more importantly: did you run your ports through poudriere-testport? it failed here and I added the missing pieces accordingly but generally it's preferred if the submitter makes sure that poudriere-testport actually passes :)

jbo: i may have forgotten that since i didn't know it existed until i submitted then - i just did make stage / make install with DEVELOPER=yes

well 'forgotten' is not the right word there, but you know what i mean

yeah no worries - it's all good :)

I am still learning too - obviously.

the problem with not using poudrier is that you might be missing dependencies etc. for example, I had to add ${PY_SETUPTOOLS} and wheel stuff. your test builds worked because that was already present in the system. poduriere-testport builds in a clean jail so it catches those problems.

does anyone remember when were all complaining about the freebsd survey the other day (meena was that you)? emaste⊙Fo says he wants feedback on it

s/were all/we were all

i really should set up logging in my irc client so i don't forget all this stuff

there are more important things than logging IRC in my opinion :p

autolog_path = "~/.irssi/logs/$tag/${tag}-${0}-%F.log";

that reminds me that I really want to check out irssi. I am still using horrible hexchat

rtprio: i think you use a different irc client than me

this is irssi

*** Client: ircII EPIC5-2.1.12 (Commit Id: 2078) (Internal Version: 20220615)

oh wow, haha

logs have certainly helped me recall something someone said many months ago

unixwitch, first two ports are commited now. I didn't look at the others yet that depend on those. I'd appreciate it if you could make sure that those ports build cleanly with poudriere-testport first.

jbo: i'll check now

meena: omg who is this igalic person on github nitpicking my commits, what a terrible person

unixwitch: when do you sleep??

meena: what's that

i usually go to sleep around 4-8AM and wake up around 12-4PM

(currently 2AM here)

unixwitch, +1

hello from the same time zone

3am here :D

i've tried to have a normal sleep schedule but i just can't do it, i always go back to this

yeah fuck sunlight :p

i usually goto bed around 00:00 and fall asleep agility after, but holidays have needed my rhythm up a bit

jbo: fascist island (UK)

lovely

oh, that was a PM, reading is hard

Also, my very good patch got merged, so i updated my ports patches for cloud-init. For FreeBSD & OpenBSD

meena, the python netlink stuff?

you use cloud-init? on freebsd?

unixwitch: i program cloud-init

well, i use it too,m

meena, ah, you work on "upstream cloud-init", not "just" the freebsd stuff?

yes

and no, the netlink stuff is still not merged

meena, that's nice!

meena, that's less nice :(

unixwitch: i live next door, glitch.social/@meena/111698011930126007

Title: eena meena me: "my two step plan for improving Ireland" - G L I T C H

oh god i have to set up a new poudriere ports tree for testport

so many commands

unixwitch, poudriere ports -c; poudriere jail -c; poudriere testport; :)

it's more all my ports are in different checkouts... i should probably fix that though

i thought that would be a good idea when i started doing this but it turns out it's not

aah

so what I do is usually having a manually managed ports tree (outside of poudriere), then just using branches for new port(s) and adding that ports tree to poudriere

this feels like a silly question but how do i download a patch from bugzilla? bugs.freebsd.org/bugzilla/attachment.cgi?id=247266&action=diff

Title: Attachment #247266 for bug #275947

275947 – [NEW PORT] net/py-pychromecast: Python module to talk to Google Chromecast bugs.freebsd.org/bugzilla/show_bug.cgi?id=275947

oh, 'View'. that was obvious

yep

and then just git-am that

pro tip: that doesn't invoke the pre-commit hooks :p

(for obvious reasons)

okay hang on, now i need to work out how to save a page in qutebrowser

just copy the URL and fetch <url>

:p

okay, this works: ilythia /s/ports-development (main)> fetch -o- 'bz-attachments.freebsd.org/attachment.cgi?id=247266' | git am

Title: Invalid Attachment ID

wut?

wut?

ah, the %27 fucked it up for me, sry

irc client parsing crap...

all good - your approach is what I had in mind.

i did try right click > save but apparently Blink saves it in some sort of weird quoted-printable format? no idea what's going on there but i will mess with that later

omg testport is running

oh i see this builds everything the port depends on since it's a different tree

unixwitch, check the poudriere config file. options PACKAGE_FETCH_BRANCH and PACKAGE_FETCH_URL allow it to fetch packages from existing binary repositories if they happen to match what it would be building otherwise.

does that work on -current? i'm not sure how up to date the package builds are here

especially useful for things like rust etc. which might take a long time to build

meena: sounds good (regarding upcoming status report)

regarding sleep cycles, I need to improve my sleep quality but my sleep apnea gets in the way :P

voy4g3r2: cgit.freebsd.org/src/commit/?id=6c9…b37170f1fb2ae8b4827070743e61b6eaed2

Title: src - FreeBSD source tree

wow that was real quick

awesome sauce

The simplest solutions are often the best ones.

yeah and even my zfs-snapshot periodic issues are resolved :) bsd.to/2t5K

Title: dpaste/2t5K (Plain Text)

jbo: py-chromecast (pr#275947) is fine in testport

py-pychromecast i mean (what a silly name)

good ole py

any reason why python ports that already have "py" prefixed to the project name can't be truncated out since we already use "py-" by convention, e.g. "py-chromecast" instead?

kenrap: i wondered about this but it makes more sense with a package name like "py311-pychromecast": this is "pychromecast" for "python 3.11"

if it was py311-chromecast, you might wonder why there's no "chromecast" package on pypi

I see, thanks

yeah, for the ability to install the same package for multiple python distributions

some have it suffixed

well loglevel = default:debug for afp.conf is very noisy

ls

Password:

apparently audio/sublime-audio depends on gcc12 *and* rust *and* llvm15

this proliferation of compilers is frankly getting ridiculous

[main-development] [2024-01-05_03h11m57s] [parallel_build] Queued: 296 Built: 198 Failed: 0 Skipped: 0 Ignored: 0 Fetched: 0 Tobuild: 98 Time: 00:11:48

296 packages for a fricking music player

I'm seriously considering getting a dedicated build server for my workstation and rpi systems

i tried cross-building rpi packages on amd64 for a bit, but some packages (like rust) are broken, and it wasn't really that much faster

obviously the solution to this is to replace my desktop and my server with arm64 boxes, but they're still so expensive

What kind of arm64 systems are you looking into?

for a desktop, something comparable to my Ryzen 5800X3D would be nice. for a server... i don't need much CPU but i guess 32-64GB RAM

Oh, yeah, that would be hellaciously expensive for arm64

yeah. if you want to spend $10k+ there's loads of options, then on the low end you have a bunch of stuff like rpi or the windows dev kit, but there's really nothing in the middle

i do have a spare M1 Mac Mini (since i gave up on macOS for desktop) which i will turn into a server as soon as freebsd gets the required device drivers :-)

Nice, or to put it in more modern terms: Noice!

unixwitch: is it just the cross compile overhead? since amd64 -> arm64

figure there is added "garbage" having a different cpu being the host verus destination

voy4g3r2: no, poudriere cross-build runs the entire build environment (jail) under user-mode qmenu

so the overhead is huge

s/qmenu/qemu/

ah hence.. have a seperate box, so it is not "killing" the current box

it's more that even on a relatively fast amd64 cpu, it doesn't end up being much faster than doing a native build on a slow arm64 cpu

and 200+ packages along with 3 compilers does not help

oh, this is just a normal amd64 build

the arm64 comment was in response to kenrap talking about a build server for rpis

ah

i have a love hate relationship with my pis

i am using one for remote backups and zfs stuff

one thing that is very frustrating with the pi and freebsd, the random "freezes"

no matter if i was directly plugged into it or over a network but they are solid toys

it's so I can isolate any build threads/jobs away from my primary system and continue to enjoy my system without interactive lag getting in the way. I thought I would also have it do arm64 builds for my rpis as well.

also considering setting up a bouncer for one of them

I got one of those argon cases where you can use an nvme stick as storage with it

weird and random question when you refer to ELF objects, do you refer to them as elf or ELF?

i ask as i am trying to go through these manual pages.. the manualpage does .Xr ELF 3 but the filename is elf.3.gz

i have it "dinged" as a mismatch do to case sensitivity

voy4g3r2: the Executable and Linking Format is an acronym, so i would say ELF, but i would type "man elf" because i don't expect manpages names to have capital letters

yeah that is where my head is at

so if it is saying ELF object blah blah .. keep as ELF

but if it is .Xr ELF 3 it should be .Xr elf 3

because that is the convetion for everything else.. happens with NUMA NewFs and for all things Sendmail

yes.. that is the way because .Xr elf 5 is right just not the reference to 3

elf(5) has the lowercase version in .Sh NAME so that seems like the appropriate choice for .Xr as well

as does elf(3)

yeah even the filename is lower case

Title: progress report – what has been addressed – FreeBSD Bug Hunting

it is just crazy how hard this is

i'd be less concerned about the filename and more about what the page actually says

imagine you printed out the manpages and put them in a ring binder, you don't care about the filename, right? because you can't see it

unixwitch: bug 108980 is all about broken links in manual pages

108980 – list of missing man pages bugs.freebsd.org/bugzilla/show_bug.cgi?id=108980

so as i am trying to fix that bug.. i am going through and documenting what i am doing

so either some of gets rejected for whatever reason or they accept it all :)

good ole man vacation (had no idea this existed till 30 seconds ago)

jbo: new patch for #275959 (audio/sublime-audio), works in testport now

jbo: and new patch for #275962 (multimedia/mediaelch) which now builds in testport, i think that's everything

god i'm so bored of vnet jails. jails stuck in dying, epair interfaces randomly disappearing...

hm, a bunch of nfs fixed just landed in main, i wonder if this fixed mmap over nfs

Is "gateway_enable=YES" absolutely "net.inet.ip.forwarding=1" and nothing else?

got my reply here; freebsd-questions.freebsd.narkive.c…om/FXkIipAD/packet-forwarding#post3 - thanks anyways.

Title: packet forwarding

 

hey everyone, so we have moved on to zfs send | zfs recv to move some snapshots and experiencing some behavior i can not explain: I setup 2 users on both machines, called backups, gave them both permission to send/receive datasets through zfs allow and chown the directory on remote side. I have seemed to forgot to do an zfs allow permission which stopped the receiving machine to get the data. I fixed

that by adding 'create' but now when i try to redo the transfer.. the sending machine says the dataset exists on the remote machine. When I do a ls -lsa on the remote machine, the dataset DOES NOT exist.. has anyone seen this and if so, what would be the recommended solution to "fix" the ghost dataset snapshot: bsd.to/NYNQ

Title: dpaste/NYNQ (Plain Text)

use zfs to check the existens of the dataset, as it doesn't have to be mounted and then ls will not see it

ahhh

so when i fixed the issue.. i get that must destroy them to overwrite it

and since it is a readonly snapshot, it is like.. that's nice try to destroy and if i did not give that user the destroy ability.. zfs goes no bueno?

well it is letting me send after a zfs destory -f, lets see if works this time, thanks nimaje

i do have to say this is much quicker than rsync magic

hi all

hi adilix

using bhyve under FreeBSD 14, I wonder do we still need the "-w" switch (the 'bhyve_options="-w"' line in vm conf file, to ignore unimplemented MSRs), for let's say OpenBSD guests/VMs?

Is there a way to mitigate Terrapin vulnerability via config options?

Oracle Linux is. . . not exactly being timely with upgrades.

CrtxReavr: try

Title: security - How do you mitigate the Terrapin SSH attack? - Unix & Linux Stack Exchange

would spiped ssh mitigate

netatalk service is awfully chatty

tankf33der, I am simply curious... Are you actively noting a Terrapin attack against SSH on your site? Note that the OpenSSH upstream stated the impact of the attack was limited. openssh.com/txt/release-9.6

Hey everyone. FreeBSD's DTrace does not implement cwd, any idea on how to get the cwd of a process?

procstat -f pid has the cwd in there

isley did it that way, it's so weird what I'm getting here

this program keeps telling me that

[ coolwsd ] ERR #16: Failed to bind to Unix socket at [snap.hell.coolwsd-39xgqStZ] (ENOENT: No such file or directory)| net/Socket.cpp:1095

but I created the socket file manually in it's cwd, and it keeps complaining

maybe it chroots or chdirs, IDK, but what would be visible in procstat, right?

voy4g3r2: Well, it is called netaTALK

vkarlsen: this is true, still can not figure out why the time machine backups jsut stop working.. i figure with all this debug stuff, i would be able to... i think it is on mac side.

well zfs send | zfs recv are operational under a backup user.. now to only get the backups themselves NOT run as root

welp, acpiconf -s 3 just shutdown the PC instead of inducing a sleep state, is that normal?

ah t says so in the handbook nevermind

nmz I've always used the 'zzz' command