where can I found the log of #freebsd ?

I don't think there's a public log. Why do you ask?

CrtxReavr: thank you again. "accept_rtadv" is the key to solve the no ipv6 address problem of my wireless connection.

V_PauAmma_V: I remembered who told me that argument.

I see curl finally got updated in the quarterly packages

RhodiumToad: You around?

_xor: ya?

RhodiumToad: Want to help me make some sense of a truss (or ktrace) dump?

ok...

One minute, let me termbin it.

Does termbin automatically add line numbers? I'll add those first if not.

Oh wait, it's plain text on there isn't it? I think I was thinking of pastebin that does that.

if you want something a little less minimal than termbin, it's easy to paste to dpaste using curl

No biggie, just ran it through nl and then nc to termbin.

That one is truss, because that's the default in my head, but if you prefer ktrace, I can do that too.

ok, what are you looking for?

Basically, I can't exactly tell where the start of the error event occurs exactly since I'm sure the rest of it is the trap/handler/whatever cleaning stuff up and dumping the message.

I'm not sure there's any error here to see.

It seems like it's somewhere around line 80ish?

Ah hmm

So here's the basic gist of what this is...

I mean I can see it writing what looks like an error message

but that error doesn't seem to be the result of any system call failure

That port I mentioned the other day that I've been using for a while, sysutils/rtx, I have an internal version of that port (since it didn't exist in the tree at the time). I just bumped it to the latest version which is failing to build. Ok, no biggie, nothing new.

There's the actual error from the poudriere testport log.

well the error in the truss output is a different one

Line 22 is the point of interest since the "build-time information: Os" is basically it saying that it can't determine the OS. Ok, great, so a compile-time generated file that embed build-info is screwing up. I suspect it's because it's missing a FreeBSD branch in some compile-time conditional.

The truss output is from a run on build-script-build, where poudriere says the build bombed.

It's the executable referenced on line 16 of the poudriere log.

... so? it obviously did something different

By "it" do you mean build-script-build?

I mean whatever you ran truss on did something different to what happened under poudriere.

There's some more context, as I dropped into the builder jail and manually ran that executable. That's the error causing the build to fail (though I'm wondering now about environment variables that are set during build, so that message might not be accurate).

don't you see that from the difference in the error?

note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace <-- did you try that?

Yes

and?

I did both, RUST_BACKTRACE=1 and RUST_BACKTRACE=full

Let me get those, one sec...

Do you know if truss passes environment variables on to the child as-is or should I run /usr/bin/env as the command passed to truss?

it shouldn't clobber the env

truss will pass the env

what would 'urdlck' status in top mean ?

sounds like a wait channel

yeah, ./kern_umtx.c:3193: error = umtxq_sleep(uq, "urdlck", timeout == NULL ?

so it's sleeping in _umtx_op, which is generally our analog of the linux 'futex'

used to implement thread sync primitives

looks like it's trying to lock a read/write lock in read mode

Title: last pid: 65521; load averages: 12.58, 31.13, 18.67 - Pastebin.com

I'm running a heavy usage web server, php-fpm has about 2000 threads running when this happens, but the upper limit is 3000

then all of a sudden the load spikes up, apparently I run out of space - hard to say, I can't even ssh to the box when that happens

Here's the truss dump with RUST_BACKTRACE=1...

stderr and stdout

_xor: that's still not the same error as in your poudriere log

You're referring to NotFound vs. NotPresent?

seriously, what part of the error message is _not_ different?

compare "thread 'main' panicked at 'Failed to acquire build-time information ..." with "thread 'main' panicked at 'called `Result::unwrap()` on an `Err` value ..."

if you can't reproduce the same error, then you're just wasting time trying to investigate the wrong thing

Dude, c'mon, I'm not a genius, but I'm also not a complete moron. NotFound is from the build tool invoked as part of the cargo build process. NotPresent is the actual source of the error, it's just not making its way up into the poudriere log.

Ok sorry for wasting time.

last1: sounds like you're running out of memory, have you looked into that?

I wanted to confirm it just now, but yeah, NotPresent is from the child process and the primary error. The parent process handles the error, but for whatever reason, it swallows it and throws its own NotFound error (I think because it's checking for the file the child process is supposed to produce, but fails to do so).

so then you need to trace the child process

That's what that was.

this is where ktrace -i comes in handy

btw, are you trying to run the command from outside the poudriere jail? that's not likely to work well

enter the jail instead and run it from inside

yeah, I don't know. I think php-fpm processes are accumulating and not dying off or something until, yes, I do run out of ram

even though I've set them to quit after serving 20 requests

It's all from within the builder jail.

_xor: the fact that your paste has a current dir name starting /poudriere/data/.m/13-amd64-main-default/ref/wrkdirs/... suggests not

You're right, my bad, this one specifically is from the host. All of the other ones are from within the builder jail.

that'll work fine for relative paths, but for absolute ones it'll look to the host, which won't work well

I'm going to delete this one and re-run it from within the jail again, one minute.

When I say all of the other ones, I mean from last night. Just now I was trying to be quick and realized I could do it from the host, but the chroot slipped my mind.

I'm just going to kill this directory of logs and start clean. Going to trigger the build again, so it'll take a few minutes, but I'll have the most recent, proper logs.

Does dpaste support uploading multiple files in one request?

not afaik.

I'm going to run `ktrace ... ./build-script-build`, any specific flags?

(already ran truss)

-i

ok

Any specific flags to kdump?

no

so, still not the same error message

Ok, how about this specific trace though? Forget the poudriere log for now. At what point in the trace did the error occur?

it occurred inside the program, nothing to do with any OS calls

Oh handled completely within the program.

gah

all one can tell from the trace is that it's before the point where it starts writing "thread '"

Yup, I originally wondered if it was because it wasn't able to read /etc/malloc.conf

and probably not much before; the preceding mmap call is probably part of a successful memory allocation

nah, that's normal

no, malloc.conf is not normally expected to exist

it's for debugging or tuning malloc by providing options

i still think the concept of malloc.conf is totally funky, but I understand it's probably the best way to do it

Yeah I saw that in the man page, but what actually reads that. Would it be the program or some platform lib?

(one of those options is to log all mallocs to ktrace)

Yes, I thought it was odd that it used symlinks :|

it's read by the malloc library, inside libc

It made me wonder if it's something that the program was trying to read and NOT ignoring the returned error from readlink, or if it was being read by some sys lib and was being ignored as expected.

Sounds like the latter.

kern.sched.cpusetsizemin also confused me for a second because it wasn't showing up on my desktop, but then I realized that my desktop runs 13.1-STABLE and the build machine is 13.2-STABLE, which is where I guess that was introduced.

that's used internally in libthr

Well, at least I have some more info now to go further. I tried patching the references to the relevant rust code last night, but that didn't work. That was when I dropped into the builder jail to poke and prod around, namely with build-script-build, but wasn't sure where the error was exactly, but didn't realize it's being handled completely by

the program, not in response to some failed syscall.

_this_ error is internal to the program. we can't tell from the information shown what the error shown in the logs is from.

Whenever we have a hardware failure during a running FreeBSD system, what log file reports the failure event furtherly/should we check first?

a failing hdd, for instance.

usually /var/log/messages

kernel messages end up there unless you configured syslog.conf in some unusual way

output of `dmesg` may also be relevant

RhodiumToad: Thank you.

i forget who i was talking to, about my chromebook - it uses i2c drivers for touchpad, and sound, and presumably network too since that didn't work OOTB; is that something i can fix by porting linux drivers with linuxKPI

ixmpp: I've seen Val do lots of work around i2c: reviews.freebsd.org/search/query/9trYNzO6ROhe/#R

Title: Query: Advanced Search

looks like the query is base64 encoded

decodes to garbage though

i think i've tried val's existing efforts in past attempts, but fair enough

Any idea why the ping acts like this?

[FreeBSD-root@:~]# ping google.se - ping: UDP connect: No route to host

[FreeBSD-root@:~]# ping -4 google.se - PING google.se (172.217.18.99): 56 data bytes

default of the ping request is UDP?

you have ipv6_prefer specified in rc.conf?

it doesn't do that for me, and i don't have anything ipv6-related in rc.conf :)

yuripv: Contrarily, I actually have: ip6addrctl_enable="NO" - ip6addrctl_policy="ipv4_prefer" - ipv6_activate_all_interfaces="NO" - ipv6_network_interfaces="none"

ip6addrctl_enable="NO" <-- i guess that is the problem

if you don't want ipv6, simply do nothing.

I don't really remember why, but I think there were services/commands prefering IPV6 by the default and I think, back then, I solved it like that way..

yuripv: that was the reason, i guess.

it's ICMP or ICMPv6, not UDP

debdrup: So, the issue isn't really the ip6* settings in my rc.conf?

btw: I've PF running as well.

*btw,

no, it's ipv6 settings in rc.conf; for me it's: $ ping -6 google.se -> ping: UDP connect: No route to host

no firewalls and no v6

"get the source address. XXX since we revoked the root privilege, we cannot use a raw socket for this."

(that's why UDP in error messages)

CrtxReavr angry_vincent meena So its a known issue and I should wait for its resolution? Also, if I install obs-studio, it removes neovim. What can be done here?

sorry, i do not follow

Sorry, gotta focus on the stuff I actually get paid for today.

angry_vincent this I mean. forums.freebsd.org/threads/freebsd-…3-2-and-virtualbox-ose.88747/page-3

Title: Solved - freebsd 13.2 and virtualbox-ose | Page 3 | The FreeBSD Forums

Beladona: my question was about custom kernel

angry_vincent what was the question?

[20:39:41] <angry_vincent> Beladona: do you use custom kernel?

angry_vincent I never made a custom one. I don't even know what it is. I just followed docs.freebsd.org/en/books/handbook/cutting-edge/#makeworld

Title: Chapter 25. Updating and Upgrading FreeBSD | FreeBSD Documentation Portal

to solve pastebin.mozilla.org/4BsTybXP/raw

why are you so against compiling that damn module with MAKE_JOBS_UNSAFE?

it literally means: it's unsafe to compile this with make -j>1

while this is true, i think, the prblem, is that this module needed be propely built against newer kernel

i do not know what is exactly different between kernel 13.1 and 13.2 changd that makes this module to barf

meena what exact command i need to pass with MAKE_JOBS_UNSAFE=YES ? at what step to use it?

Hi everbody, i make customs in freebsd's source and now how to make a ISO based in freebsd source?

Beladona: you can set this in /etc/make.conf ro maybe with `env`

but i do not know if such variables can be set once or not

@angry_vincent is possible make inside linux ? or only inside another freebsd ?

i do not understand the question, sorry

@angry_vincent Sorry, I don't know if I made myself clear, my English is not good enough to write, come on. Is it possible to generate a freebsd iso using a linux distro? Or will I have to install a freebsd (or another BSD) to do this?

mbnunes: freebsd builds on FreeBSD, Linux and macOS

mbnunes: i do not know, i have not tried building FreeBSD outside FreeBSD

Beladona: you set it in make.conf

Ever open your irc client.. scroll back a little and just decide "Nope. there's too much dumb there today" and close the client? Thats about where i am today

meena I thought its like `make MAKE_JOBS_UNSAFE=yes`

or env MAKE_JOBS_UNSAFE=yes

if anything, it would be make -j1

meena isn't it -j1 by default if someone do not metnioned `-j`?

I am doing make make MAKE_JOBS_UNSAFE=yes buildworld buildkernel

it is not about buildworld and buildkernel

it was about the module itself

angry_vincent then what should be done?

echo "MAKE_JOBS_UNSAFE=yes" >> /etc/make.conf

try building virtualbox-ose-kmod again

angry_vincent building? like this? `cd /usr/ports/emulators/virtualbox-ose/; auto-mark-install-from-source emulators/virtualbox-ose-kmod kbi `

no idea what does this do

just make install

angry_vincent `cd /usr/ports/emulators/virtualbox-ose/; make install` ?

Yes.

fckin freenode... if you try to connect irc.libera.net instead of irc.libera.chat it sends you to their servers

ops sorry, wrong channel

Hi, how do i fix this situation? I upgraded matrix-synapse to the newest version and now it's broken because py-canonicaljson was upgraded to the wrong version. see: bugs.freebsd.org/bugzilla/show_bug.cgi?id=270737

Title: 270737 – net-im/py-matrix-synapse: 1.80.0 fails to start with devel/py-canonicaljson 2.0.0

Is it possible to downgrade py-canonicaljson?

if you have an older version in your package cache: sure

Oh great, someone is going to do a custom build of FreeBSD and call it FreeBSD. Yeah, that's definitely going to go well, considering it's trademarked.

what where?

y'all made me go on "hacker" "news".

meena, You go so we do not have to

this is the first time in years I've opened the front page

rtprio: backlog.

meena: why would you willingly go there?

madetrarked

FrisBeeSD (why just infringe on one ™?)

i only see a question about building freebsd iso with modifications; don't see anything wrong with it, i do it all the time to test stuff

did i miss a java jre w/o x11 somewhere?

angry_vincent done. Now what to do? reinstall vritualbox?

debdrup when do you think obs-studio and neovim conflict be resolved?

Beladona: maybe, i do not know. if virtualbox-ose is fixed you can try rebuiling virtualbox too

angry_vincent not fixed. Same issue

ie pastebin.mozilla.org/4BsTybXP/raw

Beladona: you may want to read reviews.freebsd.org/D39474

Title: ⚙ D39474 FreeBSD Handbook: OS upgrades: kernel modules

angry_vincent I tried again and this time kldload vboxdrv just worked

I guess the rebuild was successful.

thanks !

strike while the iron is hot

:)

?

ok, so what's the syntax to properly set root's shell to /bin/sh? chsh -s /bin/sh root didn't take across logins

root shell is already /bin/sh i believe

but is simple chsh -s /usr/local/bin/bash while logged as root or as user

( it was example )

Demosthenex: "didn't take across logins" — yeah, no, it wouldn't

huh?

why wouldn't it?

btw, probably not a good idea to change root's shell to something other than sh or csh, I'm pretty sure it still gets invoked unnecessarily by the 'service' command

RhodiumToad: logged in sessions aren't gonna change shells

RhodiumToad: good grief

I assumed they meant it didn't stick across logout/login?

'service' and rc.subr generally are a crawling horror

so many levels of shell quoting and parsing

if it weren't for the fact that all the important files are root-owned and not writable by others it would be a security nightmare

obs neovim bugs.freebsd.org/bugzilla/show_bug.cgi?id=267656

Title: 267656 – x11-wm/enlightenment multimedia/obs-studio: luajit-2.0.5_6 conflicts with luajit-devel-2.1.0.20221004_1 on /usr/local/bin/luajit

anyone know what to do about this: bsd.to/mQym ?

Title: dpaste/mQym (Bash)

trev: looks like it's expecting some directory to exist, which doesn't?

yeah maybe the profile, but that does exist. maybe it's looking somewhere else

the "can't find profile directory" error may be a harmless red herring

eventually i get "Firefox is already running, but is not responding. To use Firefox, you must first close the existing Firefox process, restart your device, or use a different profile."

you would think it would just open a new tab...

oh wait i didn't notice the "not responding"

i wonder if it's cause i don't have dbus running

RhodiumToad: Figured out what the issue was, need to implement a bit of code to fix the build. Question last night was more an exercise in reading the trace, that's why I wasn't as concerned about the poudriere log. I got the general idea from the trace, but I wasn't sure if there was anything that jumped out (e.g. uncommon return value from a

call) that I was missing.

truss/ktrace/etc are only useful when the error is being induced by something that the program is getting from the OS

(or not getting)

Right. I couldn't see anything obvious in the trace, but I'm not confident enough in my reading of it to assume that there wasn't. I get that it didn't fault in there and that the unexpected state was in the program and also handled by it. Just wasn't sure if I was missing something obvious in the trace, hence the more experienced eyes looking at

it.

Last time I dealt with traces, it was on Windows. I've not done much of anything on FreeBSD at the system level, though I've been reading /usr/src little by little.

I have a pretty weird situation. Running FreeBSD 13.2 and I have an Intel X550-T2 nic, it's 10Gb

FreeBSD uses the ix driver for it, but on the first port ix0 I can only get 1Gb/sec for autoselect

if I plug the same cable, from the same switch into ix1, I get 10Gbps

in rc.conf I have just ifconfig_ix0="media autoselect"

no speed preference whatsoever

is that two ports on the same card, or two cards?

my first guess would be not getting a good enough electrical connection on all necessary pins in one of the ports

two ports on the same card

any idea when freshports.org/textproc/libxml2 might get an update to not be vulnerable?

Title: FreshPorts -- textproc/libxml2: XML parser library for GNOME

when is libxml2 not vulnerable

excellent point

so it's like the jack itself that you think is faulted somehow ?

could be

the higher speed you want to go, the more important it becomes to have good connections everywhere

weird. I did see that the ix0 port has two lights: green + orange, ix1 has: green+green

last1: Is it because of speed duplex differences?

it's not

I suspect the card is faulty as Rhodium said

there's quite a few hits on Google of people having similar issues

I always held the Intel cards in high regard - but can you guys recommend something better for 10gbps ?

meena: across logins = new connections

a new login shell should have spawned, and it was ignored.

Demosthenex: does it look right in vipw?

Title: zpool list -vNAME SIZE ALLOC FREE CKPOINT EXPANDSZ FRAG CAP - Pastebin.com

~280Tb online with a single dev special device

gotta love zfs

283T 972K, heh

that does not count as "online" (yet) :p