Hi, I am trying to do to do a make buildworld of the releng/13.1 branch without modifications on a PowerBook G4 12" (powerpc) and it segfaults during compilation. bsd.to/bb9T This happens on 13.0-RELEASE and on 13.1, meaning it's not possible to upgrade from source without reinstalling.

Title: dpaste/bb9T (Plain Text)

What additional info is needed for a bug report and how do I get that? I tried getting a trace with gdb and lldb but was unsuccessful trying to make them automatically attach to the right processes.

Has anyone else tried upgrading from source on powerpc? Or has this never been tested on actual hardware and I am the only person in the world insane enough to wait ~30 hours for a build to fail? :P

And in case someone wonders: 867 MHz single-core 32-bit PowerPC CPU (7455 rev 3.3), 610 M RAM.

bendodroid: you don't have to wait 30 hours

rtprio: what do you mean? I meant a fresh compile takes up to 30 hours for the error to occur.

how much swap do you have?

2G

I'd say a stack backtrace on the coredump file would definitely help. See docs.freebsd.org/en/books/developers-handbook/tools/#debugging if you need an example for getting one.

Title: Chapter 2. Programming Tools | FreeBSD Documentation Portal

make NO_CLEAN=yes NO_CLEANDIR=yes buildworld will rsume from where you were

rtprio: Oh, seems I was missing the NO_CLEANDIR option when I tried to continue. Will try that.

it won't start at the exact statement but it should save quite a bit of build

V_PauAmma_V: The problem I had with getting the trace is that when running make with lldb/gdb I didn't find a way to make them attach to the new process when make execs cc etc. and to automatically return to the make process when cc exits successfully.

I can use "set follow-exec-mode new" to automatically attach to the new process when make execs cc, but once cc exits, I didn't find a way to have it automatically re-attach to make to then attach to the next spawned compile job

That's why I specifically suggested "on the coredump file".

Oh, I see. My rc.conf contains dumpdev="AUTO". I was under the impression this enables coredumps to /var/crash but no coredumps appear there. What should I do to get one?

Look for a "llvm-tblgen.core" file with a recent creation timestamp. That's the coredump file for the process. (dumpdev is for panic kernel dumps.)

Ah, makes sense. Is there a central directory for the coredumps? Or are they put in the processe's working directory? Which would mean either /usr/src/ or /usr/obj I guess?

The working directory, I think.

Ok, thanks for the help and the tips. I think I can continue on my own for now :)

You're welcome.

bendodroid: /var/crash is for the kernel crashing, not processes

It would be swell if "pkg" would produce a single "port currently does not have a maintainer" note for multiple packages instead of one per package #wish_list

hey all, I am logging on via ZNC on a remote server, nothing interesting there ... but the remote server has ipv6 and ipv4, so I set: ip6addrctl_enable="YES" ip6addrctl_policy="ipv4_prefer" in rc.conf; but even so ...

01:17, * [daemon] is connecting from *@2001:41d0:801:1000::1e9e 2001:41d0:801:1000::1e9e

it appears to be ignored

hey daemon

hello ClaudioM

does znc not have something like a '-4' ?

possibly, I also could just tell my nameserver to refuse to return AAAA addresses

but I would have though setting a preference like that in rc.conf would somehow inform all applications to try ipv4 first

i assume you `service ip6addrctl restart` already

yep

that being said the reason I do not want ipv6 for irc is that last time I used it, it was a bit ... unstable i.e. connection drops every few hours

maybe its improved

that was back on freenode and all

ah ha there is a workaround for znc though, I can tell it to bind 0.0.0.0 for outgoing connections; by default it will try :: then 0.0.0.0

hello :)

do you know if it is possible to run X session from shrc or bashrc files? Let me explain this question.

I wrote a dialog script that shows after tty login (no login manager). The script asks what to run, icewm, i3, tmux, sh, bash. And I can't run X sessions from this script (Display or X errors). After this, script ends, tty show up and I can type startx succesfuly.

what may be a reason of these X errors?

What error do you get? Something about need to start X11 system with escalated privileges?

no. Timeout about connecting to Display.

( Hunh )

i use functions "icewm" and "i3". These functions are included into shrc.

.oO( Wonder if login process is not complete ) How do try to start X11 session from those functions?

so the commands are just "icewm" in tty and dialog eighter.

icewm do: startx ~/.icewm/etc/xinitrc

~/.icewm/etc/xinitrc loads ~/.xinitrc

and "icewm" works fine when typed from hand

i think dialog script doesn't work because shrc didnt finish yet.

Yeah, that is what I am thinking

so the script probably should be a system service exactly like login manager :|

At that point just as well use some display manager

i will search for something easy to configure

maybe some tty login manager

easy to configure because I need use my functions

You also have the option to turn the functions into shell scripts

yes

it is an option

but first, login manager must be easy to configure to use exactly these script to start x sessions

will think about it ;)

thanks for talk ;)

🤝

btw: I wonder if it is possible to ask for firefox-91

it would be great to have 91 version in repo

version 91 works fine. Newer are slow.

i had to update to newest, because 91 stoped working after upgrade

Create a VM of older version of FreeBSD, install there from ports

i cant compile firefox

i use 32 bit machine and compiling firefox require 64 bit. Another reason is that I use old computer.

VM of some old OS of whatever persuasion seems to be the only way from here

hmmmm

old computer + vm = slow old firefox ;)

Oh I forgot one thing that without CPU support for virtualization, thus in software, VM in general will run tremendously slow

Sorry

What is the best place for gnu?

Not here

Really?

I expected this big thing about hating on linux and the real gnu supporters are bsd which is freebsd?

(im here to learn obviously)

nothing wrong with Linux

I happen to like Debian quite a bit

i don't hate linux, just systemd and networkmanager

yes i hate systemd its so verbose and esoteric

rtprio: I run Debian on my workstation. No systemd, no networkmanager, no pulseaudio.

i dont have a lot experience with nix but really interested love the command line and its been over a lot of years though i started in fedora fc4 i think it was and back then idk i dont remember systemd from back then init scripts and runlevels were the things to learn anyway

systemd and pulseaudio ;)

oh no ... i was trying to learn on a chromebook for a while and audio was so bad

sadly, freebsd isn't free from pulseaudio

seems like you have to have kernel level forensic knowledge

mason: debian without systemd? Isn't it Devuan?

There is #freebsd-social channel for all the non-FreeBSD conversation needs

i quit Linux when LXC begun to require systemd

so I use jails now

and jails are much better in my opinion :)

i watched and interesting google talk about cgroups and jails and namespaces really interesting stuff it made me feel like we are users are basically working the "kernel space" of yesterday almost-like

an*

working IN the kernel space of yesterday

obv didnt really know what they were talking about but all that complexity really pulls me in really

i may work in the kernel space of stone age if it works fine ;)

wikan: Since you're not in -social, no, plain old Debian: github.com/ChibaPet/install-debian-zfs-sysvinit

Title: GitHub - ChibaPet/install-debian-zfs-sysvinit: Install Debian on ZFS, with sysvinit, optionally encrypted.

wikan: Also, LXC doesn't require systemd: github.com/ChibaPet/lxc-utils

Title: GitHub - ChibaPet/lxc-utils: Tools for composing content for containers.

yet^

That said, please join #freebsd-social as I *am* off-topic with this.

it is possible to run without systemd but it is much harder than year ago

Actually... Past time for me to become a pumpkin. G'night all. o/

it is because cgroups use systemd services.

wikan: Other way around.

and docs support only systemd way of doing LXC things

im trying to be in -social now but thats really interesting to me you know like the inner circle scientist stuff from yesterday that they were really working on/with is now like a norm that i get to experience thats really cool

what is -social?

wikan: /join #freebsd-social

ohh

hello people someone can help to me i wiil try run from freebsd 13.1 the re::engine::RE2 and i get error bsd.to/E1DK

Title: dpaste/E1DK (Plain Text)

ok well i guess i scared someone away

i just want to say i love tech, tech in of itself isnt good or bad

someone please can help?

I managed to get a 6.6M kernel, stripped to 5.4M. I guess I should reboot into it now?

and I need to load a kernel module for ahci.

I have to assume kernel modules consume wired memory?

on this episode of misconceptions about freebsd

what provides the /dev/ufsid things? I built a minimal kernel and my configuration depends on this, which isn't available anymore. it seems I was able to boot by supplying ada0a at the mountroot prompt.

I managed to boot with 81

81MB of RAM

wait, how very pointless

IT HAS 8 VCPUS.. AND THAT LITTLE AMOUNT OF RAM

what the hell

Reinhilde: what machine is this??

meena: qemu.

meena: These results aren't realistic at all.

Reinhilde, "git grep -F ufsid" suggests glabel(8) is involved in filling /dev/ufsid/. What value does sysctl kern.geom.label.ufsid.enable have?

V_PauAmma_V: I figured out I needed to load geom_label, thanks (the sysctl was 404).

(as in it did not exist)

*nod*

Hey all, how are you? What's the latest version of FreeBSD kernel that you have on 12.3? it is 12.3-RELEASE-p6?

i'm doing fine, thanks for asking.

antranigv, -p6 for kernel, yes.

someone before was talking about mailcap and mutt.. when I get a html mail, it opens in mutt as text instead of calling links.. any ideas on how to get it to call links? I have the mailcap in place.

maybe mutt only detects mails as html if it's an html attachment not just Content-Type: text/html; in the mail header?

ah never mind. I had to add: auto_view text/html in the muttrc :)

hi.. php74 did eol on freebsd. there is no way to get it from packages?

Not from the official repos, no

You'll have to de-deprecate and build them yourself

so there is no archive for packages?

but ports archive?

ports, git checkout 27ac371f93d36f77f00b8da261e496904184dd33

i will get it from there

ivaat: generally, it seems inadvisable to run unsupported / unmaintained software…

so unless you're running this on a computer with absolutely no Internet access, i would rethink the effort

it is in jail and all known XSS, SQL injections are filtered in nginx

sry, just general software does not come along so fast to support php 8.x

12 nov was last php 7.4.33 update so i think like one year is ok to run it at least :)

ivaat: how hard could it be to upgrade? (n.b.: i have not followed php development since 7.2)

php.net/manual/en/migration80.incompatible.php wat.

Title: PHP: Backward Incompatible Changes - Manual

i don't how hard it is.. it is specific enterpise php software.. i am not programmer myself. i just panid and use admin mode ;)

and project must be live next week

on development they will do upgrade too

-panid+panic

-how+know

How come that clear_tmp_enable is not enabled by default?

tct: excellent question.

I need some help. Trying to use an Intel wireless device (iwm0) following the handbook (docs.freebsd.org/en/books/handbook/…vanced-networking/#network-wireless) but getting error when starting netif service: "iwm0: device timeout, dumping device error log, errlog not found, skipping"

Title: Chapter 33. Advanced Networking | FreeBSD Documentation Portal

harpia, which card?

(haha, palmface)

tct: Intel Dual Band Wireless AC 9560

harpia, you might wanna give iwlwifi a try

I run a 9560 successfully, as iwlwifi.

jup

nice to see you on irc, cracauer :)

Is this all I need to do? wiki.freebsd.org/WiFi/Iwlwifi#I_hav…vice_and_want_to_try_iwlwifi.284.29

Title: WiFi/Iwlwifi - FreeBSD Wiki

Nice to see you, too.

harpia, you'll need to run 13 for sure. personally, I track stable/13 on my laptop solely for the purpose of having the latest iwlwifi efforts available.

I'm on 13

harpia, if you go down that route, subscribing to wireless@ ML would be a good idea, I guess.

meena: that PHP 8.0 incompat table is hilarious with the '==' operator. Makes for sloppy programming. :)

pertho: I've taken this to #freebsd-social

Will iwlwifi eventually replace iwm? Just trying to understand what they are.

harpia, man iwlwifi(4)

> iwlwifi will be a successor to iwm(4) and may superseed that driver in

the future. It still complements the iwn(4) driver which supports older

chipsets.

it's not a successor, iwm is based on native code where iwlwifi transforms linux code

also bear in mind, these days, a lot of heavy lifting is done in the 802.11 stack with the firmware

Is their an archive of older 14-CURRENT ISO image snapshots? none of the once on the FTP site are currently bootable on my system.

What boot problem do you have? Many can be fixed with bootloader settings.

Boot up halts during the hardware probe: venus.morante.net/downloads/unibia/…91/14-CURRENT-4dcc6baddb4/boot1.txt

I forgot to save my good kernel, lol

My plan was to either grab a bootable one from a more recent snapshot, or build one using the live CD

How old do you need? Maybe I have one.

Anything after July 4 2022 (commit hash: ec32fc2af52) or before what's on the FTP site. Worst case is I try to boot from some older ISO's that I have from early last year.

I have Jan 4 14-current right here.

it's aarch64 btw

oh

no worries, I'll just boot of one of my really old ones and try to rebuild my good kernel again.

I had wondered why the dmesg looked so strange.

There are a lot of Decembet-January images here: ftp.freebsd.org/pub/FreeBSD/snapshots/arm64/aarch64/ISO-IMAGES/14.0

Title: Index of /pub/FreeBSD/snapshots/arm64/aarch64/ISO-IMAGES/14.0/

Yeah, I tried as far back to Dec 9

some change between July 4 and Dec 9 broke something

I am going to try and track down the commit

It should be possible to build a kernel using a live CD env, right?

Never tried. But I don't see why not.

What machine are you on that is broken?

it's an ARM64 server: R281-T91 made by Gigabyte Computer

Very cool.

I've been messing around with it for the last few years. FreeBSD hasn't been too stable on that hardware. I figure I would provide as much feedback as I can to help out

I'm sure it is much appreciated. Attractive machine. And apparently quite cheap.

just reading up on that hardware, seems very nice!

Wait, froogle display the wrong machine. Dunno what it costs.

It wasn't so bad, it was about 3k from where I got it. But it's normal cost is about 5-6

3k is an awesome price for that.

otis, ping

meena: I'm back at this bridging problem, I'm exploring potential options where just the main host is not acting as a gateway properly between the two bridge interfaces

meena: for example, I created a new epair and added that to bridge0 and bridge1 to see if other epairs on bridge0 would be able to route to bridge1, no dice yet

the route table on the host looks correct too, Basically the jails host can ping into any host off bridge0 or bridge1

rtyler: bizarre

i wonder how small the most minimal reproducer is we can construct

a number of articles I'm digging up are referencing vlans, which I know I wasn't using before, such as: bsdrp.net/documentation/examples/multi-tenant_router_and_firewall

Title: Multi-tenant router or firewall [BSD Router Project]

we can just use / or rescue as path to demonstrate

I also tried to set up a port forwarding rule in pf, and I cannot go from jail-on-bridge0 to postgres-in-bridge1 via the external interface, even though another machine can get to the postgresql port through that external interfacfe

rtyler, is this hitting bugs.freebsd.org/bugzilla/show_bug.cgi?id=268717 then?

Title: 268717 – [pf] rdr rules don't work for traffic originating at localhost

maybe :shrug:

rtyler: it feels very steep to have to use a vlan to do this basic a thing

agreed

I feel like vlan is for people who didn't sleep thru networking class in school

so watching tcpdumps there's nothing hitting bridge1 when I'm trying to go from a jail off bridge0, I'm going to reconnect my epair between those two and see if that changes it

meena: watching bridge0 it looks like the router for bridge0 jails (i.e. the host) isn't responding to arps telling who has the bridge1-based device

do the two bridges have different MACs?

yes, gist incoming

meena: gist.github.com/rtyler/08e3179a68b8831317f87b040b04da8c

Title: Some details from the jails host · GitHub

sounds like we have i bug

wait just one stinkin' second

nope, I was misreading my output, dangit

meena: welp, where to now boss :)

I have been thru this before, and doubted my sanity. but now we have three people failing at this, so i say we create a minimal reproducer and submit the bug

rtyler: you can also use more descriptive names, and IPs that are easily told apart, so people reading the bug don't go thru the "wait just one stinkin' second" process

sure maybe, gotta get my bugzilla account back, and also spin up a VM to create the repro case while I move this jail over to the one network :(

I can try, after im back home and my daughter is in bed

we're… late

heh, I'm assuing you're CET

PST here, so lots of time to wait around for buzilla ^_^

rtyler: I moved tu Ireland recently

absolutely

going to try to reproduce this now, gives me something to stream abouht

cracauer: do you have to rebuild the kernel to make it work? wiki.freebsd.org/WiFi/Iwlwifi#Rebuilding_Kernel

Title: WiFi/Iwlwifi - FreeBSD Wiki

No, iwl on my 9560 worked with the default 14-current kernel.

rtyler: you stream?

meena: presently yes twitch.tv/agentdero

Title: Twitch

not regularly

in my test environment I am pinging across these bridges, wtf

rtyler: love it. absolutely thrilled with this turn of events

quite perplexing, adding more jails now trying to find the changed variable

V_PauAmma_V: FYI: Your tip to look for .core files was correct, I found 3 make.core dumps and just opened the bug at bugs.freebsd.org/bugzilla/show_bug.cgi?id=268974 Thanks for your help :)

Title: 268974 – make buildworld fails with SIGSEGV on PowerBook G4

Nice.

I strangely enjoy being in #freebsd-bugs, which as VimDiesel post all the bugs that are created

meena: it seems to be related to the subnets I've chosen

with bridg0 is is 10.10.1.1/24 and bridge1 is 10.10.2.1/24, then it doesn't workj

bridge1 is 192.168.100.1/24 (for example) the routing works properly

How do you persist this across reboots? wiki.freebsd.org/WiFi/Iwlwifi#I_hav…vice_and_want_to_try_iwlwifi.284.29

Title: WiFi/Iwlwifi - FreeBSD Wiki

Something like this? if_iwlwifi_load="YES" in /boot/loader.conf. And probably if_iwm_load="NO".

rtyler: that's still a bug, but now we've reduced it a lot

meena: definitely! I'm *super* perplexed by this

meena: what's your blog domain again, trying to find your jail post to share

rtyler: alpha.pkgbase.live/howto/fresh.html is not a blog

Title: Howto: Setting up a Fresh Install

meena: ah, thanks

it's funny how my advice of: make it easy to read led to discovering how it make it work

I'm trying some different netmasks now to isolate where the bug occurs

meena: I think I have a theory

meena: when bridge0 and bridge1 share a /8 the routing doesn't work

meena: when bridge1 is on a different /8, it routes properly

this has been an interesting debugging experience @rtyler :D

rtyler: that's confusing because you don't use /8 in any stanza

z3uS: always fun when it takes several days to pin down

meena: basically if bridge0 and bridge1 are both in 10.xxx land, no routing

meena: bridge1 in 172.x or 192.x works fine

@meena lulz

oof, been slacking too much don't need @'s on IRC ;)

rtyler: im gonna need to

meena: after I write the bugzilla, I'll link you

look at the code, because that sounds like someone is worse at math than i am

network math.

z3uS: my client just gives me autocompletion for names that way in a nice GUI drop down

<3

I should really stop using kvirc with my znc at some point but I'm too lazy

I'm soon going to follow dch's reviews.freebsd.org/D38058 and port Thelounge.chat

Title: ⚙ D38058 www/hedgedoc: NEW PORT web-hosted markdown doc editor

meena: bugs.freebsd.org/bugzilla/show_bug.cgi?id=268976

Title: 268976 – Traffic will not route across two bridges on the same /8

z3uS: since you were hanging about too ^

bendodroid, you're welcome. And tangentially, "find /usr/obj '*.core' -print" saves you a pipe and (I think) some typing.

V_PauAmma_V thank you! <3

V_PauAmma_V, you need a -name before '*.core' ; and on the other hand the -print could be omitted

daemon: cc bugs.freebsd.org/bugzilla/show_bug.cgi?id=268976

Title: 268976 – Traffic will not route across two bridges on the same /8

-print can be omitted, since ca 1989

(rough estimate)

heh

Always having -print is always "safe" though if you want it to print. Because it depends upon the find rules that come before it.

I sometimes get snagged by running a complicated find rule and then nothing gets printed. I look and then curse myself out for driving it wrong.

jilles, oops.