yeah, that's a bit of a corner case because we basically never expect someone to have freshly installed a -p* build

why would it be so odd for someone to install an actually up to date build?

because we, for better or worse (no, definitely worse), can't re-roll install media for a release

so even if there's something critical enough that it really should be rolled in to the release media, that just won't happen right now (though re@ could probably figure it out given a bit of time)

people rolling their own release media in the wild basically never happens if you look at the nonexistent pile of reports that come from that scenario

so really the only supported way of deploying freebsd is install a potentially insecure version of the OS then rush to perform network updates before getting rooted

embellishing a bit but still, the model is broken imo

yup pretty much

doesn't feel great on the user side fwiw

ootb there's not a whole lot of options for remote accessibility in the first place

huh?

polyex i think ootb is out of the box

this is another one of those overly-abstract problems, out of the box there's very few vectors one has to exploit remotely and ideally you're not silly enough to start deploying services before you've updated it

i don't even know what remote accessibility means

if you take off the 'ibility' maybe

eh still tho, to require network connections to freebsd servers to bring a server from installed to up to date sucks

I'm done with laptops for today (possibly) lol later all!

later jb1277976

it doesn't require network connections to freebsd servers necessarily

if you're that paranoid you're probably running your own freebsd-update server in the first place

no the whole idea of installing an out of date OS is what doesn't make sense. the network update after install step should be totally optional

feel free to file a bug report

your point isn't unreasonable, but almost every mainstream OS works like this, unless it has a way to include patches in the install media (like Windows does), but even then hardly anyone actually does that

re is under new management so it's a great time to revisit things like this and try to understand why the current model exists and what can be done to improve the situation

but none of that will happen in #freebsd unfortunately

how can i debug why sudo pkg upgrade outputs pkg: an error occured while fetching package?

it doesn't say which host it's trying to connect to lol

you know, it's odd that a chip called "Killer E2500 Gigabit Ethernet Controller" doesn't actually support 2.5Gbps Ethernet. could they have chosen a more confusing model name?

lol

logically it should support 2.5 terabit ethernet

unixwitch: every time i see that, my dyslexic brain makes this out of it im.eena.me/uploads/1a53e1b4f920d7f4/ethernetkiller-4112566390.jpg

There is no brace expansion in sh as there is in bash... how do i replicate this behaviour in sh 'file{1..100}'?

for i in `jot 100`; do echo file${i}; done

Anyone ever received this error? <jemalloc>: jemalloc_rtree.c:205: Failed assertion: "!dependent || leaf != NULL"

I don't quote know what it means or have remalloc_rtree.c in /usr/src

rtprio thank you!

fikran: src/contrib/jemalloc/src/rtree.c

no idea where the jemalloc_ prefix comes from in the assertion output

success! 64 bytes from 10.1.4.1: icmp_seq=0 ttl=64 time=0.222 ms

except i'm not sure how i fixed it, i just set the interrupt mask to ~(uint32_t)0 ... probably need to work out which particular bit it wants

kevans: thank you! Curious why its mislabeled...

kevans: what does this error meaning? I don't quite get it from looking at the code.

fikran: no idea, sorry

jemalloc and i haven't been friends since they ripped out the redzone support that was really useful

weird, I'm not even using jemalloc

i think it's the default allocator, i.e. if you do man malloc, you get the jemalloc man page

yeah...not sure how to proceed heh.

This error makes no sense to me.

yeah, jemalloc is the only libc malloc in freebsd today

do you think FreeBSD devs would accept restructuring sys/dev/usb/wlan/STUFF to sys/dev/usb/STUFF ?

I can't really imagine a good reason to do that, no

usb tries to be reasonably organized by function, why would you want to merge all of this?

can anyone see if there is anything wrong with this zfs output? bsd.to/fLxN

Title: dpaste/fLxN (Plain Text)

when i try to connect a client to the server.. it keeps giving me this tcp] 192.168.0.120:/storage/blinksystem: RPCPROG_NFS: RPC: Program not registered

when i try to mount the drive.. which DID work at one time

kevans: to be consistent with other USB drivers

fikran: which other drivers? i feel like you're missing that the vast majority of usb drivers are in functional subdirectories (net, serial, video, storage)

everything left directly in sys/dev/usb are mostly infrastructural bits

voy4g3r2: i fought with something like that for a while. try rebooting. the issue is sometimes that the rpc crap gets all registered at once, so if you restarted rpcbind, it drops those registrations. sometimes you will get lucky (it was inconsistent for me), but at some point i realized that i had the configs correct and just restarting allowed all the things to start in the correct order that the nfs shares registered

hrm.. at this point might as well give it a try

scoobybejesus: i know that nfs option works.. i have used it before

well looks like i will be right back

you should see it with showmount

hrm.. still happening

showmount on the server gives the expected output?

-m? -u? I forget

and then showmount <server-ip> from the client

i get that RPG message again

showmount: can't do mountdump rpc

from the server, you get that error?

yeah

you don't have this folder in /etc/exports, right? zfs sharing over nfs is supposed to be easy (I think)

I double checked my notes. you don't need anything in etc/exports, and you don't need to reload mountd or restart nfsd service

nope.. i have no /etc/exports

Title: dpaste/BaqS (Plain Text)

this is my /etc/rc.conf

messy but i think i got what i need

i know services are working from zfs.. because netatalk is now running.. through time machine on my laptop

you just zfs set sharenfs='network 192.168.0.0/24' storage/blinksystem. you don't need to set it 'on'. try without the maproot part

same result

you have nfs_client_enable instead of nfs_server_enable

son of a gun

i don't know if you need nfsuserd_enable, and i don't think you need (but it couldn't hurt, i don't think) nfsv4_server_enable

was just going to ask that v4 is NOT loaded

add the v4 to the /etc/rc.conf line?

again, you may want to reboot because of how these services register themselves with rpcbind. but service nfsd start might do it.. not sure

it loaded :)

that v4 line is a separate line. you'd keep both

nice!

now to see why this /etc/fstab is garbage

thank you

happy to help. just realized my nfs server is not working now. i thought it survived the jump to 14.0-release. maybe not. i will look into that later

funny you should say that.. last time i fiddled with this was on 13.2

and was like.. why is this not working anymore

thank you scoobybejesus , time to get this git server moved to the zfs

voy4g3r2, In your previous paste bsd.to/fLxN/raw there is a line "on,-maproot=chrisdavidson:blinksystem,192.168.0.0/24". What syntax are you using for the 192.168.0.0/24 part? It's just hanging there in space. I have never seen that syntax before. I expect it to be something like a traditional "-network 192.168.230.0/24" or "-network 192.168.230.0 -mask 255.255.255.0" or something and not just an IP address there.

Title: fLxN

rwp: in an article i read it said take what you have in /etc/exports and mold to blah

and make sure on is there..

but i like the new way better :)

it looks prettier

That is true enough. It should be the line from /etc/exports. In which case what will the "on" do?

if I install ports at install time, will freebsd-update update my ports tree?

mmlj4, No. Ports and base are individual and separate.

Upgrade binary pkg ports with "pkg upgrade"

rwp: the one was suppose to invoke zfs sharenfs=on

so use the git pull method? or...?

rwp: I'm asking about ports, not packages

If you have ports compiled locally then you would follow the make install command for them, which I don't recall off the top of my head.

never mind, thanks

The main issue to be aware of for ports is that they compile and link against shared libs in base. When base updates some of those shared libs are obsoleted and removed. So ports need to be compiled against the new set of base shared libraries.

you're still talking about apples, I'm asking about oranges

Otherwise they try to run, find a shared library that previously existed in base is now missing, and the port can't run.

You said compiled ports. But regardless what I said applies to both.

I did not.

I asked about the ports tree

apples, please

can anyone answer my actual question? thanks.

Sorry but I think we are talking past each other.

4.5.1. Installing the Ports Collection

Before an application can be compiled using a port, the Ports Collection must first be installed. If it was not installed during the installation of FreeBSD, use one of the following methods to install it:

Procedure: Git Method

mmlj4: the answer is no, freebsd-update won't touch a ports tree

ports tree, not compiled ports.

kevans: then the tree is orphaned?

effectively; historically you'd manage it with portsnap, these days you'd manage it with git instead

fair enough. Thanks.

is portsnap a dead project?

kind of... ports folks were insistent on deprecating it because it didn't handle some things very well (quarterly vs. latest, maybe? my memory is foggy) but the servers are alive and well probably for the foreseeable future

the maintainer was somewhat against deprecation because there are many valid uses

it seems to be working, for now anyway

the git method kept crashing for me with different errors each time :-/

with git you probably have to specifically do ssh cloning with anongit@

or clone it through github

the ports tree repo is massive and has historically had some issues trying to do HTTP(S?) cloning

if portsnap keeps working I won't need the git method, regardless

I'm down the road, thanks, kevans

dang I need a sound card, but my pci-e slot is taken up by bluetooth wifi card :/ I think

there's only one?

hmm let me double check

yeah it could have two but I think one is obscured by the width of the graphics card

it has been a while since I took it apart

I'm having an issue running a freebsd router with openbgpd using pf as my firewall. I want to be able to restrict traffic to my bgp port from only my peers, and otherwise unrestrict traffic in/out. Is this something anyone can help with?

ok so freebsd gets buggy if i try to build releng/13.2 and latest patchlevel. but is base what's buggy or is it just freebsd-update, and i can just ignore for now?

Can't say, for the latest patch-level may be buggy for you but it has not been in my case🤷🏽‍♂️

parv so you git pull a branch like releng/13.2 with latest patch, build, build installer, install with installer?

polyex, I think "releng/13.2" would be a Git "tag" not a "branch". Further, I have not notice (do update me if there are) any tags for each patch level. In any case, I follow the "stable/1[34

]" branch & pick a point in the commit history to install

you pick a point based on what, stable code?

based on guess, past mailing list messages, & specifically when "__FreeBSD_version" changes

k then you run that and don't use freebsd-update?

Right (on some hosts I use "freebsd-update" & others update from the source)

Just updated from source in branch "stable/14" due to "__FreeBSD_version" change as that would become significant in near future for package updates

ok ya. today ppl were acting like i had 3 heads because i was building locally instead of using release plus freebsd-update

was doubting myself for a sec lol

tbh i thought MOST ppl with large fleets deployed up to date code instead of installing old shit + needing a bunch of update downloads

obv not

Whatever works. When I started with FreeBSD 14, it was still -CURRENT thus would not have "freebsd-update" support. Any newer installations, of 14, would be updated via "freebsd-update". A clear advantage, & only one AFAIAC, of updating from source is that I do not have accept all the changes since last patch-level update.

ya

i'm looking at unbound's pkg cgit.freebsd.org/ports/tree/dns/unbound but where's the default unbound.conf that we get in /usr/local/etc/unbound after installing?

Title: unbound « dns - ports - FreeBSD ports tree

In one case, really need 14-CURRENT for GPU driver support. In another, I happened to have only 14 installation image, did not bother to download+copy 13 one; so.

s/need/&ed/

Re unbound, see the "pkg-plist" file; in there at the very top is the sample configuration file location

@sample %%ETCDIR%%/unbound.conf.sample

but how does it get into etc?

it would have to come from the unbound pkg itself no?

Yes

Could be in the directory wherever "unbound" is built

so current is 14.0.RELEASE-p4 - is there a URL or something where this string of latest can be found?

kolla, Outside of errata or security announcements, I do not know; there is freebsd.org/releases/14.0R/errata but does not have exactly that you are seeking

Title: FreeBSD 14.0-RELEASE Errata | The FreeBSD Project

right, I only found it there as well

i can't believe it's so hard to find the unbound.conf that gets installed with the pkg. how is that good for learning?

What? Did you not find a configuration file after fetching+extracting the source (given use of Ports)?

no web

polyex, Do you "unbound" installed? If so, try "pkg info -l -x unbound | grep "[.].conf"

s/you/& have/

ya i do

parv: releng/x.y are branches

kevans, Ah. Thanks much for the correction

kevans, While you are here, are patch levels tagged in some way too?

unfortunately not

polyex, Correction for "pkg | grep": pkg info -l -x unbound | grep '[.]conf'

kevans, Thanks

I vaguely recall asking someone about this recently, but searching my inbox for "tag patch levels" is a disaster

parv ya ty. i can do that but what i wanted was a way to browse online the package as a unit of what it'll be putting on my OS

so i could see what its default unbound.conf looks like

maybe that's not useful to anyone else

we don't really maintain a set of ports distfiles all extracted anywhere

lots of storage needed

ah

we've had requests from time to time to have such a thing for searching, though

sometimes what you want can be found in files/, but in this case and many others we're just installing something from the software's release archive as a sample

polyex, In that case seek out the project web site|page & hope that it has put up the information of interest

okie

Is anyone using FreeBSD in dual-boot with Windows 11, each on its own partition on a single disk, scenario? (There will be space+connection for only one SSD in the future laptop)

... yeah I could continue to use FreeBSD VM as currently I am ...

... but the reports of people getting ads in Windows 11 make running the VM not as appealing.

ugh, what ads

yuripv, See the results of duckduckgo.com/?t=ftsa&q=windows+11+ads&ia=web -- one from there: elevenforum.com/t/disable-ads-in-windows-11.8004

Title: windows 11 ads at DuckDuckGo

looks like fud

ok🤷🏽‍♂️

alpha.polymaths.social/@rl_dane/statuses/01HEVP2KV9ZX6WKX0SVFJD7PNX reminded about running Windows in a VM instead!

Title: Post by rl_dane (@rl_dane⊙aps)

... but the laptop (HP EliteBook 640 G10+|Lenovo ThinkPad T14 Gen 4+) hardware would have to be properly supported by FreeBSD for me to that even consider🤦🏽‍♂️

i can type resolvconf -a vtnet0 < /path/to/file to pass the file contents in, but what if i wanna pass 'a string of text' in instead of file contents, how should the syntax be?

Use "cat" with "heredoc": termbin.com/57b6

8 lines to pass 1 string into resolvconf?

That was just an example. Do not know|care what or how "resolvconf" needs

Last 2 lines are the output of "cat | awk"

There is also "echo" or "printf": printf "%s" "whatever" | resolveconf ...

... "echo" has different usage than "printf"

hmm after an update to 14.0-RELEASE sudo and pkg where both defunct - needing libcrypto.so.111 and libsso.so.111 - fortunately it turns out they where elsewhere on the disk and I had user elevation though su

now I have a nagging feeling I have a halfway upgraded system

any way to rerun the installation for 14.0-RELEASE again ?

not libsso but libssl

BarnabasDK: openssl was upgraded to version 3 in base; you should rebuild all dependent ports (or force-upgrade packages)

i think freebsd-update tells you to do just that (if you used that)

yes - I saw that in the release notes. I just think it is strange that the upgrade of the OS deletes the needed old shared objects before you get to this step

so - I copied in the old libs - did the upgrade of pkg and a reinstall of sudo - and deleted them again

you had pkg-static for that

ah ..

there is a statically linked version - did not know that

BarnabasDK: did you upgrade with freebsd-update or from source?

freebsd-update

BarnabasDK: so when you do a release upgrade with freebsd-update, you run install three times. the first time installs the new kernel, the second time installs the new userland, and the third time removes obsolete files, including shared libraries. you need to update packages/ports to the new release between the second and third freebsd-update install runs to avoid this.

anyone familiar withj nis?

there a command to append some text to a file? echo "foo" > file.txt doesn't append it overwrites

polyex: echo "foo" >>file.txt

yes!!!

wanna be fancy, you use tee

added bonus, it works with sudo

so this github pull system seems to work quite well, and i imagine it's good for attracting contributors, but it's kind of a shame it depends on github

could set up an official freebsd gitlab instance?

from what I searched, there already is

any clueas why ypbind wont bind on my freebsd client but on linux it's not a problem?

Title: FreeBSD · GitLab

woo, ypbind, now there's a relic from the past :D

mane: I can think of a few possibilities... lke ipv4 vs ipv6

*like

cxan you be more specific?

are server and client speaking same protocols?

they run both on 13.2

ah, ypserv is also on freebsd?

yes

all systems on same lan?

of course

what's the error message? "server not responding"?

ypcat: no such map hosts.byaddr. Reason: Can't bind to server which serves this domain

is there a reason crypto/heimdal hasn't been updated from upstream for ~10 years? (not a complaint, i'm just curious, especially if there's something i can do help with it)

unixwitch: do you mean the port or the base libraries?

meena: the base libraries

oh that's weird

wow

it's missing some useful newer encryption types, like the SHA-2 based ones (we're stuck with aes256-cts-hmac-sha1-96, upstream has hmac-sha256-128-aes128)

meena: the vendor branch has been updated but it doesn't seem to have been merged in...

cgit.freebsd.org/src/log/crypto/heimdal the last merge was 1.5.2 in 2012

Title: src - FreeBSD source tree

So then the question is, what's the difference between that, and cgit.freebsd.org/src/tree/kerberos5 ?

Title: kerberos5 - src - FreeBSD source tree

is that the freebsd-specific Makefiles etc. for heimdal?

yeah, like kerberos5/lib/libkrb5/Makefile pulls from ${KRB5DIR} which is crypto/heimdal

ah, sorry._. I'm still just on the phone

so i install a default disc1 with basically nothing but defaultrouter and ifconfig_ in /etc/rc.conf.local. so obviously dns fails because i don't have a resolver setup yet but why can't i ping my router? lol

ifconfig shows the expected ip for the if

netstat -rn shows default with the gateway ip of my router

so wtf

My current task: find out why snmpd on jail foo reports Memory Usage, but snmpd on the host itself does not. It must be configuration. snmpd config is not just a single text file. :/

polyex: is the interface up? what error do you get from ping? what does tcpdump say?

ifconfig says the if status is active. ping says ping: sendto: Host is down

'active' isn't enough, it needs to have UP in the interface flags, like this: ix1: flags=1008943<UP,BROADCAST,RUNNING,PROMISC,SIMPLEX,MULTICAST,LOWER_UP> metric 0 mtu 1500

ah ya it's UP

does tcpdump show packets going out on that interface when you run ping?

how do i use tcpdump for this?

in one window run 'tcpdump -ni <ifname>' then in another window run ping (or whatever) and look at what tcpdump prints

k gotta enable ssh gimme a sec

it's gonna take me a bit i gotta rework some stuff. weird that i couldn't just get a network connection to my router lol

hey everyone, i started receiving some i/o errors on my zpool and the only way i was ble to clear them was with a reboot...

the zpool is now operational again but i have a strong feeling, this is NOW dying too.. is there a way to perform a health check of sorts?

voy4g3r2: Does zpool status tell you anything?

vkarlsen: it says everything is good now.. after the reboot of the machine

voy4g3r2: Have you scrubbed recently?

voy4g3r2: what was the `zpool status` before you rebooted?

no i have not

good morning everyone

morn

Title: Message ID: ZFS-8000-HC — OpenZFS documentation

it pointed me to that.. for some reason dmesg is now blank... trying to pull up the logs

smartctl is the tool to ask the drive how it's doing; sysutils/smartmontools

i am having a really weird zfs issue, which only started since i upgraded to 14.0 (which makes me wonder if it's a bug) where a pool gets random data errors, but only on snapshots (never on files) and which vanish on reboot

i thought it was a memory issue, but i've run memtest86 twice and it didn't find an problems

Title: dpaste/mAD6 (Plain Text)

that is one drive.. nothing looks out of wak

whack

oh, a WD green

yeah...

i know i know..

if i can keep these stable enough.. i want the data OFF the dang things to i can properly evaluate this stuf

i was hoping to wake up, my time machine backup would be done.. and i can start "moving" the data off

you could do a long test, but i wouldn't trust a green further than i could throw it

Title: dpaste/OQi5 (Plain Text)

it is a lot of information

haha.. yes i have read that and i have had people tell me that

Will ipfw in-kernel NAT support syntax like?: bpa.st/QCVQ

Title: View paste QCVQ

I'm specifically referring to the port ranges.

It's supposed to be a (mostly) drop-in replacement for natd, which supported them - but I haven't found any examples with the in-kernel NAT.

just type it in and see if it accepts the syntax

well ltest see what these 235 patches do for the release 14.0 updates

did not expect this many updates since release.. oh well

unixwitch: what's next on the list?

meena: maybe fix this zfs issue :-) i've given up on alc(4) for now, i mailed the Linux driver maintainer to see if they have an actual documentation for the thing...

i don't follow how it only errors on snapshots

it seems like only metadata is affected, when the problem occurs "zfs list -t snapshot" will fail to list some of the snapshots and says "cannot iterate filesystems: I/O error" ... but there's no i/o errors, no checksum errors, and all the data is perfectly fine

unixwitch: alc or acl?

voy4g3r2: alc

the driver for AR81xx Ethernet cards

ah okay.. i am going through documentation "bugs"

and i remember seeing acl issues

in this case the issue isn't freebsd documentation, but documentation for the chip itself :-)

Title: FreeBSD Bug Hunting – An inquisitive person finding their way in a FreeBSD world.

haha

bughuntingfreebsd.wordpress.com/2023/12/24/boss-we-got-a-list <-- specifically this mess

Title: boss!! we got a list!!! – FreeBSD Bug Hunting

that does sound weird

it is very weird, to the point it really feels like an issue with the hardware, but then i'd expect at least cksum errors

rtprio: would you recommend a scrub for this zfs.. before i try to get this data off as you just "love" the wd greens

what's a nice desktop font? i don't like Deja Vu, the spacing seems weird

Title: Programming Fonts - Test Drive

i have been enjoying this font

Title: Nerd Fonts - Iconic font aggregator, glyphs/icons collection, & fonts patcher

you can download there

but my days are usually in man pages, nvim and other console tools.. not much of a gui person on the desktop

i use a gui mail client (claws) since i haven't found a terminal client that handles multiple imap accounts and folders very well... maybe i should look at that mutt sidebar patch

that i just use my phone.. mailing list recovering addict here

i remember penn state telling me, i was overloading my email account.. i have been recording every since

like 20 years ago

rtprio: i know not many fans of the wd green.. what about the barracuda line of drives?

ypwhich: can't yp_bind: reason: RPC failure

any ideas what's wrong?

mane: are your rpcbind services running?

yes

rpcbind is running

does dmesg or /var/log/messages say anything more?

i was having similar rpc pains, last night, and it was a tango that i was stepping on feet

god i hate roff syntax

can we convert all the manpages to latex?

voy4g3r2 maybe you'll think something out, really need this project to work

unixwitch: you just have to breathe thru it, and at some point you don't even see the syntax any more

I don't know how long it takes to reach that point, probably writing and rewriting some thirty odd man pages, but, yeah

unixwitch: Can, yes. Although 'can' and 'should' are different words. First step would be to add tex support to man and related commands.

meena: it's giving me flashbacks to highschool when i set my homework in troff

unixwitch maybe you'll know what's wrong?

Also, picture what the base system would look like, if man supported LaTeX

probably… different

mane: with your YP? sorry, i haven't used that for ages. i use LDAP nowadays.

Very different.

maybe we'd stop pretending that our computer is a teletype, and start treating it as a computer?!

unixwitch maybe you got a good ldap tutorial for me|?

i want to know how to configure the ldap freebsd servers and clients

not sure about a tutorial. for client i use net/nss-pam-ldapd, for server, openldap, it's bit of a hassle to configure but the documentation is good

what are the criteria for marking a commit as "MFC after"? or is it just use reasonable judgement?

unixwitch: kind of a reasonable judgement thing; if it fixes something and can be backported without breaking KBI/ABI somewhere important, that's a good candidate

unixwitch: it would be nice to just be able to have al the man pages in one semantic syntax.. regardless of roff or latex.. i am trying to "find" broken links.. and there are multiple ways to make a link.. .Xr is not the only way..

mane: you are trying to get bind operational? i apologize not 100% sure what you are trying to get operaitonal.

voy4g3r2: what are the other 'commonly' used ones?

i'm trying to make ypind work

ypbind

it works on linux clients but can't bin on fbsd clients

meena: bughuntingfreebsd.wordpress.com/2023/12/24/boss-we-got-a-list in the observations section

Title: boss!! we got a list!!! – FreeBSD Bug Hunting

Pagename(section)

\flpagename\fR(section)

\fBpagename\fR(section)

also .Xr all produce links in man pages..

aaah, i've used Pagename(section) myself before, actually…

yeah, i have been documenting the "fixes" and what i am learning in the process.. there

it all started when i was trying to read up on zfs(8) stuff and a link did not work

man.freebsd.org/cgi/man.cgi?domain ← struct protosw {

Title: domain

has a bunch of that, and I didn't realize it's "official"

but i still went out of my way to correct them

then you got weird ones like zzz(8) does not exist in base install

the focus is just what is in FreeBSD 14.0. and NOT including ports

which all circles back to the zfs stuff being a pain as all my work is there

voy4g3r2: i would generally focus on what's in CURRENT, since that's where you gonna contribute fixes to

can you briefly explain why the ZFS stuff is a pain?

> RESOLUTION: Removed this reference as it is not in the base OS but in the ports

I don't think that's necessarily how we want to handle those

kevans: possible, it is why i am documenting what i was doing for the world to see

my focus is NOT ports but base os..and leave it up for debate outside of my efforts

meena: i have thought abou the CURRENT thing, then the zfs stuff was acting up

meena: i think the hard drives in the zfs pool are starting to die and i am having weird status issues, drives sometimes connecting.. all because i am using usb.. i have not figured out root cause

kevans: which one in particular are you referring too?

it would be nice to annotate SEE ALSOs with, these are in ports, installable via $package-name

that one was blk(8) reference in zfs(4) if I'm reading this right, but it's more about the general problem

I don't think we want to axe a link just because the software is in ports

blkid(8)

we have blkid in ports? i could've just depended on it, rather than writing a parser for geom?

i thought of this $package-name route then was getting lost in ports

that one's probably a bad example because I think it went away entirely in a later zfs

oh, util-linux… brrrrr

oh that's neat, I didn't realize freshports lists man pages provided separately (e.g. freshports.org/devel/util-linux)

it is interesting when you get down to things like speaker(4) if you do not have a speaker on your computer.. it does not get installed but appears to be part of base install

the new ZFS manpages are full of Linuxisms...

kevans: it's a new development

kevans: it's new.

that's incredibly useful, thanks

then a fascinating one is the whole atof_l() vs atof()

the difference is precision of holding data in a variable.. but one has a man page and another does not

atof_l() no man page.. but references in the stdlib.h

kevans: Thank Graham Perrin too. Blame him. FreshPorts/freshports #492

Title: manual pages: if a page can be inferred from the pkg-plist, then link to the page · Issue #492 · FreshPorts/freshports · GitHub

492 – sysinstall shows "success" after "no space" failure. bugs.freebsd.org/bugzilla/show_bug.cgi?id=492

voy4g3r2: it should probably just be documented in the same man page as atof()

meena: yup, the "fix" i just pointed atof_l() to the atof() manpage.. dragonflybsd did that

so we can just import it from them?

i did not do that... i just looked at what they did and replicated it manually

this is my first bug i am working through

so learning as i go :)

voy4g3r2: it's simple, you just add the dragonfly git repo as another remote, and git cherry-pick -x the commit where they make the change, then clean up the mess that creates because we've drifted too far apart.

it is why i baselined 14 instead of CURRENT.. i know for a fact CURRENT will change everyday. and that chaos just adds a variable i am not prepared for yet

the good thing about using CURRENT is that it's a lot easier to work on this incrementally if you can get finding problems down to a science of some sort

some of these will have disappeared in the interim, and it's hard to say how many that might be

well that output is generated through an awk script :)

it looks through /usr/share/man hierarchy and does it magic.

do people run current on desktops nowadays or just use a VM?

i used it on a laptop, until i bricked it

voy4g3r2: sweet, so that's pretty solid

now I'm too chicken to install it on *this* laptop, which is running some Ubuntu derivate, so I'm running it in VMs

i think i mentioned the other day that the last time i tried it was during the removal of giant locking which kind of put me off running current forever

imo the optimal approach here is to roll with -CURRENT and just re-scope it; e.g., today I'm only dealing with section 1 manpages. your patches will need to be against -CURRENT anyways, so whittling away at the problem there is ideal

i should *maybe* upgrade this Ubuntu derivate some time, that might solve some issues I'm having with it…

unixwitch: removal of giant lock? when is that gonna be?

unixwitch: i've rolled with -current on my laptop for somewhere around 6 years now

unixwitch: that sounds like a good idea for a post on this blog

a VM of current on this mac

meena: hah, well not removal, i mean during 5-CURRENT when fine-grained locking first appeared

kevans: yeah thanks, a person on X has been trying for years and was able to help with building one.. tim chase

(although... isn't giant going away in 15 or 16?)

unixwitch: it's going away any year now.

I think I've started using it full-time when VIMAGE was merged into GENERIC

giant had its tentacles in the kbd/console area, which all gets a bit sticky

in a RELEASE, so, 10.1?

kevans: good to know. maybe i'll try it (with a backup be...)

(had, probably still has)

i think atkbd and psm still use it, which surely can't be removed yet. maybe when i386 is removed...

last time I tried re-locking kbd it got hairy

yeah, you can fix atkbd to not use giant with kbd not using giant, which also requires some work in vt(4) and sc(4)

unixwitch: two words: Boot Environments

to be fair they did say 'backup be'

(i should've used them, it's the reason my old laptop is now a brick)

kevans: don't you know I can't read

meena: that's what i meant by 'backup be' :-)

unixwitch: don't you know I can't read?

(i'm so glad freebsd has caught up Solaris's Live Upgrade from the 1990s :-D)

my first job was solaris, so I thought that was *normal*, and was shocked to find out that it's not

solaris is far from normal

kevans: I've also briefly touched AIX, and

Solaris is the reference implementation of UNIX System V Release 4, so by definition, it is the most normal OS. :-)

what unixwitch said.

bah

and, like, who here hasn't built SVR4 packages?

i used to have a whole custom ports-like toolchain for building those

same…

in retrospect i probably should've used pkgsrc but i don't remember if it worked on solaris at the time

well, mine was inspired by Archlinux

unixwitch: I should've used OpenCSW, but one of the senior admins vetoed it.

"it won't be as optmized as it can be if we don't compile it ourselves"

i have wasted life-times on compiling PHP…

did they know you can't compile solaris

(well okay, you can, but like)

i felt like a bit of an arch user when i put CPUTYPE?=znver3 in my make.conf the other day. maybe that's more gentoo than arch.

unixwitch: try CPUTYPE=native and see what breaks

meena: i have not built SVR4 packages.. i did get to play with a HP-UX box in the late 90s early 2000s

btw, if you remember my setuid ping_exporter complaint from the other day, someone (not me) submitted an update to it: bugs.freebsd.org/bugzilla/show_bug.cgi?id=275628 ... hope this gets committed soon

Title: 275628 – net-mgmt/ping_exporter: Update to 1.1.0

i even got to dabble with carroll.psu.edu back in the day..

my personal web server used to be an HP-UX box... i quite liked it, even if it was weird in places. (less weird than AIX though)

the packaging system was quite nice, you could bundle a bunch of subpackages (like bin, man, doc) into a single file and select which to install

unixwitch: it's already committed, cgit.freebsd.org/ports/commit/?id=b…3d03a7dfab1f70b755b668d0feab4b40bb5

Title: ports - FreeBSD ports tree

oh, that was 14 minutes, i didn't reload the tab :-D

it's too bad they didn't structure it to fix the setuid issue then update it, but I guess it doesn't matter that much with the end of the quarter days away

kevans: imagine if we could do more than one commit in ports, you know, like in git…

'magic'

meena: you could, i suppose, attach the output of a git format-patch with multiple commits to a bug? or is that not allowed?

unixwitch: i don't know how ports committers do anything

speaking of commits, somebody review and merge my extremely good pull requests, while i go find something to eat

(and reboot the freezer)

well that is just werid.. i thougth when i would do a zpool scrub, it would just run in foreground.. nope it runs in background.. 33% complete :)

a film 20 years from now will open the a voiceover, "In a world... where people had to reboot their freezers..."

silicon valley? didn't they do that whole thing with the fridges

voy4g3r2: you *can* run it in the foreground (zpool scrub -w) but it just runs it in the background and waits for it to complete

i sort of like in background.. it will be done in 3 hours.. and hopefully i will know more about what is goign on

looking at replacement drives now.. no idea what to even get.. i just know not WD green drives.. as rtprio just loves those :)

it's really just a sensor that's broken, not a real computer issue… but I'll need to find someone who knows how to fix big electronics, knows how to use a soldering iron, and knows where to buy the sensor in question, in rural ireland.

until then, "rebooting" it is.

my partner and i both work in the tech industry, and thus, none of our heavy duty equiment has computers inside

voy4g3r2: i like the Seagate Exos drives, which are available in both SAS and SATA flavours, but they're quite expensive. you can find them on eBay easily though, often without too much wear

HGST are also good but i don't know if they make/made SATA disks

oh wow.. for a 128meg catch 4tb is it is 75 bucks on amazon

got to 8tb and 256 jumps to 175

two 16TB 445

wow that jumped quickly

unixwitch: thoughts on the barracuda ones?

i haven't used one of those for a while... from what i remember they were fine but they're designed as desktop disks so they might not like running 24/7, i had a couple die

well gives me something to think about when i gym it up and listening to some podcasts.. bsdnow.tv

Title: BSD Now

buying consumer hard disks or SSDs nowadays seems like a nightmare because everyone lies about what the device actually is

voy4g3r2: ah yes, the soothing tones of benedict and tom (and i don't think i've actually listened to an episode with jason tubnor just yet)

what kernel options do i want for CURRENT on a desktop? i'm assuming turning off INVARIANTS and WITNESS is probably okay (unless i run into a panic or something)?

What do those options do ? I've not built a kernel in ages, not just for FreeBSD, for any other OS

mns: WITNESS is for debugging mutex-related issues, INVARIANTS is code to check a bunch of things that should always be true (like asserts)

unixwitch: ok that makes sense. Yeah I doubt you'd want those for a desktop system you were using regulary

kinda depends on what exactly you're wanting to work on, I think

yeah, i guess no one is going to pay me to write sctp manpages

klara's often got at least one or two contractor freebsd roles open, they've got folks in and out of north america: klarasystems.com/careers

Title: Careers | Klara Inc

unixwitch: I would just compile both, GENERIC and GENERIC-NODEBUG, and run GENERIC-NODEBUG and boot GENERIC when you're in trouble

me, use generic? like some sort of pleb?

(personally, I always just ran GENERIC)

(this is a joke don't @ me)

(well, GENERIC-MMCCAM, which, as it turns out, it's actually MINIMAL)

i noticed your PR for that earlier, is this about CAM for mmcsd(4)? or something else

yes, just equalising it across platforms

Also, adding VIMAGE to MINIMAL, because that's pretty useful

re: this heimdal thing, i'm wondering if it makes sense to backport the new algorithms from current heimdal if we can't update it in base. does everyone just use the heimdal port? but that doesn't help with base sshd/nfs/...

why can't we update heimdal?

last full update was just in january, but I vaguely recall that cy was maybe talking about kicking substantial parts of it out of base and maybe switching the rest to mit

kevans: are you sure? the version in main is 1.5.2 which is >10 years old

the vendor branch was updated but it wasn't merged in, as far as i can see

ah

that's unusual

meena: i can't find it now but there was a post somewhere (maybe on the mailing list) that basically said "it's really, really hard to update heimdal because <reason>"... but i can't remember what the reason was...

i wonder if therer was som echange to admin-y bits in the interim and we can't update without losing some compatibility with older but supported freebsd releases

it wouldn't seen unreasonable to keep libkrb5 in base and shove the rest into ports (kadmin, kdc, etc)

yeah, the idea was that if you need a server implementation just get the whole thing from ports, but we'd keep just enough client-side bits in base to authenticate against / use kerberos

since those tend to be more compatible with either implementation iiuc

is there a list for discussing this sort of thing? freebsd-ecurity?

-arch probably

-security

ah

the s is ecurity stands for security

kolla, Release patch levels from unofficial source: bokut.in/freebsd-patch-level-table via mastodon.bsd.cafe/@grahamperrin/111658459330472909

Title: FreeBSD Patch Level Table

since we just branched 14 off this would probably be a great time to try and spur movement there

unixwitch: go go go!

i would be happy to push on this since the current situation seems terrible, but i don't want to annoy people

maybe i'll start by asking a question on -arch :-)

annoying people is the only way to get things done

yeah, ask there and cc cy@ specifically just in case

(until you get a commit bit, and then you're seen as the blocker and people start to annoy you)

if you're interested in doing it, getting consensus and making sure you're not stepping on cy's toes are basically the start

meena: i spend far too much time drunkenly computing to get a commit bit anyway

i bet the src part of this is relatively easy and straightening out ports USES=gssapi is the annoying part

kevans: it is not bad with him.. but i ahve been only really listening to them since episode 500

hot new idea for C34 or C45: make __result_use_check the default

voy4g3r2: oh yeah, I'm sure he's good... just kind of ran out of time for podcasts myself

kevans: get rid of your dishwasher, if you want more time for podcasts

i think I'd have better luck getting rid of my eyedoctor / dentist

but she's too cute to evict, and the surprise dental inspections have really calmed down in the last 12 months

kevans: yeah i only get time when i head down to the gym.. it is part of my get out of the 2019-2023 mess.. i use to gym it up 5 days a week.. my body is like what are you doing to me man

hrm.. this current install through virtualbox is not as bad as i remember

current is not as unstable as many people make it out to be

yeah i even went the route of bootonly image amd64 and letting it install on the mac

this will help isolate the manpages work.. because i "guess" i should do this work with current over release 14.0

heck even using an image ont eh zfs.. might as well use the dang thing

loader(8) seems *really* slow on a 4K display, and i'm pretty sure it's because the method it uses to clear the screen via EFIRT is really slow, and it does this like 3 times

i wonder if it could be fixed to just not do that

well if you want to send me over that 4k monitor.. i could take a look

no promises though :)

voy4g3r2: you wouldn't want it, it's either broken or about to break...

although weirdly since i switched from macOS to FreeBSD on my desktop is it working a bit better

this VG2732m-LED from viewsonic is showing its age

yeah well going from a retina to a non-retina display "whatever the heck retina means" is hard

i couldn't go back to 1080p now, i use one sometimes in our home "office" and it's so ugly

i had an apple display.. back in teh day a 23" cinema display dvi version.. that was nice

i sold it as i had no room, at the time, for 300 bucks

which was not bad at all.. now this 2012 15" retina is holding on for dear life

well that is just fun.. installer finishes with a checksum error

recently made the horrific discovery that with three displays in my house, the only 1080p display was our old 55"-ish TV

horrific because the gear I was working on wouldn't boot with a 4K efifb, and neither of the 4K displays let you fake EDID to get 1080p where I need it

no wifi on it either, so had to relocate it across the house to debug its boot failure then move it back

oh no, graphics/drm-kmod is broken on main

oh... or it's trying to use src from /usr/src for some reason...

iirc you have to set SRC_BASE for the port build if you're using a different location for your tree

admittedly haven't built a kmod port in ages, I just cloned the drm-kmod rwpo locally and LOCAL_MODULES it in so that it gets built as a tied kmod w/ buildkernel

actually i think was confused about /usr/src, this is in poudriere so it should get the /usr/src from main

voy4g3r2: i had both the 23" and 30" cinema

they were great

yeah they were

pricey but worth it

hopefully this install of current works.. the latest 20231216 bootonly is giving checksum errors

could be that i am using netatalk to transmit the iso images.. we shall see.. because kenny loggins danger zone

netatalk? what the shit

yeah, why would anyone choose AppleTalk when you could use DECnet Phase V?

rtprio: because it was arleady setup for the time machine .. maybe samba? this is all happening on a mac book.. wanted to make it easy :)

i did setup samba for my music library streaming :)

okay, well maybe i should move this image.. that is 2 images from a file server that were not aligning checksum

ah, fixed it, just needed to update to latest ports tree

well okay.. bootonly does not work on a virtualbox.. time to just try disk1 instead

is the llvm FORTRAN compiler really called 'flang'? this just looks like an obscure swear word to me

god flang it

okay thi sis just odd.. how can 3 different iso files from current all have checksum errors

how are you downloading/copying them? don't say netatalk...

haha

well wget

to get them

i even tried installing directly from machine, same result

i am going to try a virtualbox image now

yes.. i have intertwined netatalk in there :)

if you run the checksum multiple times on the same machine, do you get the same result?

voy4g3r2: because iso newer or older than latest snapshot

they don't get updated at the same time?

if all three are from the same build that probably makes sense if we're fetching one or more set

what checksum, i just downloaded the iso images and went from there.. am i suppose to also download the checksum file?

oh, i thought you were downloading the ISO and then checking the checksum of the ISO you downloaded... but you mean the checksum error is during install

yes, it is during install.. my bad

but now my timemachine backups are acting up also

let me reboot this laptop and give this another try

voy4g3r2: not all dist sets are on the release media, some have to be fetched remotely

iirc the checksums for all are stashed on the media

if you have an older snapshot and have to download a dist set, boom, blows up because dist sets are replaced with each buils

build

(along a branch)

unixwitch: do you use New Flang, or Classic Flang?

I'm getting this error for a non bootable SSD. "GEOM: da2: using the secondary instead -- recovery strongly advised." should I ignore it?

s2r: does it contain a ZFS filesystem? i've had that when using zpool create if the disk already has a GPT label... (seems like a bug)

The device has zfs. pool: zssd - ONLINE --errors: No known data errors

in that case i would just ignore it. *don't* use gpart recover because it will wipe the ZFS label and replace it with whatever was there before

I haven't noticed this error in previous reboots. This time is running 13.2p8

nice how the Dovecot docs for setting up Kerberos tell you to do everything via samba-tool, like no one uses Kerberos except on Windows anymore

kevans: Do you know if there's active work going on for the Bluetooth stack?

_xor: pretty sure there's not at all

:/ I was wondering if I was missing something other than blued.

unixwitch: I don't do windows.

tmp_: neither do i, i was being a bit sarcastic there

Indeed. I felt it called for an obvious followup.

although i did run into an actual problem, which is that dovecot seems to have no obvious way to map kerberos principals to userdb users...

i guess you need to do that in ldap

Windows is a fun little game launcher.

Not something I've dug into. I work with systems with ldap, kerberos, nss and dovecot.

I don't play games, I wonder if that's why windows and I don't get along

i spent so much time playing /usr/games/larn many years ago. apparently that's been removed from base at some point :-(

oh all of /usr/games was removed?

darn those fun-hating freebsd developers

s/darn/flang

is it bad practice to put comments in manpage sources? i feel like this would make them a lot easier to navigate when editing

Some things from /usr/games moved into /usr/bin

factor, rot13, primes, morse.... But hunt is gone. I ran that one a bit.

unixwitch: if comments are useful, then comments should be put in the man page