BaloneyGeek: which version of FreeBSD are you on?

meena: FreeBSD 13.1, fully updated

servus

dch: is that hungarian "hello", maybe?

meka: its Austrian hallo, or maybe more Viennese!

More like Barbarian^H^H^HBavarian.

Rhabarbarian

cracauer: while linguists put Viennese under "Bavarian dialects" aussiehst

*Austrians would not accept that

I see their lebahkaassemma and raise them an eitriger

WE ARE NOT THE SAME

My chromium build finished. 9:15 hours. Not too bad.

is that with the most up to date ports tree?

As of yesterday, yes.

Happy that it even built clean.

haha i hear you

i'm going to give it another try then

I also made a package (amd64) in case that is useful to you. It has pulseaudio support, though.

ty but i need to build it myself. lol it's personal now! :D

I know the feeling.

Hi. Is it possible, there is some sysctl property that may cause segfaults, traps and system error interrupts on many applications?

i can't find why so many apps dont work at all and puke trap or segfault messages

for example: falkon and otter browser

Did you try lldb on them?

hmmm lldb is better as I see

signal SIGSEGV: address access protected (fault address...)

so it is a "trap" as I understand

the same error on another app

no, not the same: invalid address

so, is it compiling error? should I report it?

Is that a native FreeBSD program or Linuxulator?

or it is because I didn't setup something, maybe sysctl

native

Then you should try to get a backtrace with lldb.

Did you compile it yourself or did you install it with pkg?

it is pkg

next application (netsurf): invalid address

Let me install otter and see what I get.

i didn'tcheck otter

hi. how do we find out how which packages are manually installed?

i checked falkon

falkon-qt

i386

(not a dependency)

falkon-qtonly

kristall, netsurf (works, but segfaults after clicking on any link)

netsurf-gtk3 works for me on amd64.

i think it is compilation problem

You say you are on i386?

yes, true

what is the name of the demon/program which receives mail? not filters into boxes like procmail neither provides imap download like dovecot; simply listening on a port for incoming mail and put it into some spool/mbox

and these errors are not only on one machine

I can't reproduce those segfaults on amd64. Are you sure your hardware is stable?

well i am not sure, but these error I have on two machines

oh

You should definitely go through the trouble of getting a backtrace out of lldb.

isn't it require debuging symbols?

not for basic symbols in backtraces.

ok, I have no idea how todo this yet ;)

i did "strace" and it didn't tell me anything

which program do you want to debug first? Falkon?

is it the same as backtrace?

falkon is ok

lldb `which falkon`

r

[qait for segfault]

bt

hmmmm nice

first one is last call, right?

i quess because of "*" symbol

*?

first line starts with *

so I quess it is last call and position of error

oh. Yes.

so the error is related to intel driver

i915_dri.so

Can you use some paste service to make the backtrace available?

oh

You are screwed :-)

let me check other apps

again intel

I just started Falkon on my Intel graphics laptop, amd64. It works.

i use old machine

I would say this is worth opening a bug report for. Especially now that you have backtraces.

definitely legacy hardware

Should still work.

Title: Simple recipe for backtracing with LLDB? | The FreeBSD Forums

thank you

Anytime.

the same driver works on both machines

i will report it

As a brutal workaround you could use the Vesa driver for a while.

yea, but vesa doesn't suppoer xrandr

and I use external monitor

Yeah, it kinda sucks.

If you have a serial port you can use a vt100 :-)

is it possible there is some dependency problem?

like no mesa libs or something?

No, that would give you a clear error message.

ld error,right?

Yeah.

So here is a crazy workaround: ssh -Y localhost

Then start falkon over the ssh connection. That will make it unable to use 3D rendeering which is probably where you crash.

nope

A bug report it is then. Did you try running on the internal display without xrandr?

no

will try

before reporting

The xrandx thing might be important here.

i will try something stupid

let me try run falkon with both screens turned on

because I turned off internal screen

yea, didn't work

I would still try on a freshly booted machine's internal display without any xrandr happening at all.

yeap,will try

How is pkg supposed to deal with global python updates? I have pkg update pull in new py37 today. But afterwards removing all python 3.7 stuff works, so that was not because of dependencies.

is there an easy way to have sendmail forward all mails to my local user?

mhh, okay, so i found out about virtusertable and added an alias, but when a mail to that address was seen by sendmail (seeing the maillog entry for it), running "mail" doesn't show anything for the user i set the forward to…

Any files at all created in /var/mail ?

/var/mail/phryk exists, but is empty.

also a bunch of other files in there, mostly named after services.

The root mailbox should have the daily run outputs.

"all emails", meaning "any local part" is a grand way for having spammers discover "known active accounts", resulting in a deluge of spam

From default cronjobs.

jgh: this is a desktop machine behind a nat. i doubt that.

cracauer: mhh, /var/mail/root doesn't exist. mail when run as root says no mail. this is on a desktop machine, so it's not constantly running, but gets >1d uptime quite frequently…

Yeah, there should be cron output in root's mailbox then.

ah yes, years ago I apparently added sendmail_enable="NONE" to rc.conf because this is a desktop. guess i'm trying again after throwing that out.

Try echo foo | at now + 1 minute

as root

umm, so many emails.

is this now throwing output of every cronjob ever run into the queue?^^

(i did service sendmail start)

I don't know. I don't think so as it shouldn't have any mechanism to queue mails with sendmail=NONE

yeah, i removed that beforehand. :P

and yes, something like that but it finished within a minute or two. also a bunch of returned message mails and postmaster notifications.

mail as normal user still says no mail, even though maillog now shows "relay=phryk@localhost" for the mail in question…

Anything in /var/spool/mqueue?

yeah, 10 files with some super-terse but patterned names.

They are mails not yet processed.

There is a sendmail command to list them along with the reason why they were not delivered.

sendmail -bp

operation timed out on the remote server handling the original mail address. i don't want that server even being contacted, but the mail to go to my local user instead. is this doable without major hassle?

also thanks for all the help thus far cracauer :)

Does `telnet localhost 25` connect and print sendmail's banner?

220 aether ESMTP Sendmail 8.16.1/8.16.1; Thu, 22 Dec 2022 18:10:24 +0100 (CET)

Out of ideas for now.

not completely sure if that's what you meant by banner, but it does reply, yes.

Yeah, then I don't know why it doesn't just connect to localhost when sending mail to localhost.

i assume it would send the relay after the original and since the original goes to a remote server on the web that's then contacted?

s/web/net/

Yeah, but it should be hard to configure a global relay by accident.

Mail config is like an endgame boss of sysadmin.

I have postfix and dovecot deployed on another machine. I think even a second set on that on my homeserver. I just had a look at sendmails submit.cf and oh goddess, no…

You really want to mess with the m4 versions of sendmail config files.

asking a friend who told me he likes the sendmail config. thinking it takes a particular kind of person to decipher these arcane incanations.^^

full disclosure: I'm an Exim dev. Opinion: there are reasons for the figures here, and the .cf syntax is one: securityspace.com/s_survey/data/man.202211/mxsurvey.html

Title: Mail (MX) Server Survey

I have ld: error: undefined symbol: environ and I see with clang -v that /usr/lib/crt1.o is not included during linking phase but I have no idea why...

Which software?

cracauer: I am building pulseaudio

let me try

cracauer: my sendmail friend delivered, got it working :D

Hm, works for me on amd64.

phryk: what was it?

Hey there, I'm planning on installing FreeBSD on one of my computers, a laptop, and I've skimmed through the handbook, and I wonder if firefox in the quarterly package repositories is kept up-to-date?

cracauer: I am sure it works fine using ports but I am building pulseaudio 16.1 using pkgsrc

Ah.

A simple example exhibit inclusion of /usr/lib/crt1.o as the first crt in the linking process

I don't know what could cause a linkink without crt1

Dooshki: quarterly packages stay the version number they are, unless there's a security update; that's the entire point of the quarterly packages.

debdrup: Understood. The reason why I'm asking is that for firefox in particular, the updates that they do release typically count as both security and feature updates (unless it's the ESR branch, which is security updates only)

So is my understanding correct that the regular branch will not be updated, but the ESR one will?

phryk: gist.github.com/igalic/c77ed494e102977c9fd06ce9b053cda0 /cc cracauer jgh

Title: disable sendmail on FreeBSD, completely. · GitHub

Dooshki: you also can go with latest

Dooshki: why not go look at the quarterly ports to find the answer to your question?

Will do. The practical answer as to why I haven't done it yet is that the only computer I'm stuck with is really slow due to havy disk I/O in preparation of backups for this reinstall :D So I was wondering if someone who uses the quarterly ports and a very popular application would know the answer off the top of their head

I am looking at it in a min.

cgit.freebsd.org is a webui, and typically quite fast

Waiting for git checkout :-)

Good to know!

At the time of this writing quuarterly and main are both on 108.0.1

Although main has a portsrevision of 2

I can reproduce the error with -Wl,--no-undefined but building a shared library with UND environ doesn't seem like a problem to me

Though, I am not sure...

I think updating firefox more often in quarterly works because there is little that depends on firefox. So you don't break chunks of the tree with regular updates.

Thanks for letting me know, that's really encouraging news! I'll go look at the logs to see if it's a long-term trend

It is hard to imagine that FreeBSD would include insecure versions. So if a software only publishes mixed security/feature updates then quarterly == main.

Yeah, that makes sense, thanks!

I'm really excited to get back into FreeBSD after about a decade, now that I'm no longer forced to have certain pieces of proprietary or linux-only programs on my machine! (I've graduated from college, and I have a dedicated work laptop at my job now - no need to pollute my own machine)

I'm seeing some pretty neat new additions, like ZFS native encryption, the pkg tool, or even wayland compositors!

I have no idea how to show you, but I have pretty bad buffering using the proprietary Nvidia drivers. I used nvidia-xconfig for the config file and loaded nvidia-modeset. Mouse is handeled by moused (probably not relevent). I am on herbstluftwm using picom (that could be an issue), but I did try xfce and same stuttering (I did enable dbus). The stuttering is happening no matter what really is on screen, it just freezes for a second

every few seconds. I have no idea what log to send because I have no idea of the cause.

What does x11/xvinfo say?

That looks good. Is top showing high CPU usage for either video player or the X11 server?

Nope everything at around 10% running a invidious video.

Which video player?

Just in firefox

Same thing with qutebrowser

Does video work fine in vlc or somesuch?

Well it's not just the video. Just moving windows or even texting freezes. Just not the mouse pointer.

Another in `dmesg`?

Nothing. I'm out of ideas. Maybe /var/log/Xorg.0.log?

On tty1 I do get some ignorable errors... but they are ignorable I guess.

18:33 <cracauer> It is hard to imagine that FreeBSD would include insecure versions. So if a software only publishes mixed security/feature updates then quarterly == main. ⬅️ software is constantly full of bugs and the only thing we can do is fix it when it's discovered, or help discover / try to prevent holes

tty1 is warning me about a UTF-8 character I didn't install I think.

or rather tty0

Guess the solution is to switch to AMD :D

Only if you have an AMD card that is neither too old nor too new :-)

I have a GTX 1660, it really is a good value card. The one that is closest to the price is the RX 6500 XT, but no codecs and still 50€ more expensive :(

Wait..... could my encryption cause the problem?

It shouldn't slow down performance like that.

I think I will try GhostBSD or MidnightBSD.

So thanks for trying cracauer :D

Anytime.

Hope you figure it out.

hello!

Compiling a shared library with -Wl,--no-undefined produces an undefined symbol when using environ

I try the same test on NetBSD with clang 15 and it compiles without error

I don't know what is the expected behavior...

triaxx: paste as much output as is reasonable

into bsd.to, obviously

just pipe to | termbin.com 9999

or is it .net

...

$ dmesg | nc termbin.com 9999

yep .com

Title: dpaste/ou2x (Plain Text)

The C code is just:

extern char **environ; int foo(int a) { if (environ) { return 1; } return 0; }

Compiling without -Wl,--no-undefined works well but adding -Wl,--no-undefined produces error

If I understand well, environ should be resolved at run-time

$ readelf -s /lib/libc.so.7 |grep environ

2: 0000000000000000 0 NOTYPE GLOBAL DEFAULT UND environ

Is it normal?

triaxx: try adding stdlib.h

you shouldn't need it

Can you post the complete commandline that fails?

termbin.com is wonderful, right

The C code: bsd.to/voFN

Title: dpaste/voFN (Plain Text)

The Makefile: bsd.to/ErF7

Title: dpaste/ErF7 (Plain Text)

The command that fails: make libfoo.so

Sorry, I should change the file type...

triaxx: I wonder if it's some confusion related to this, noted in the Debian environ(7) man page: bpa.st/M6OZI

Title: View paste M6OZI

Otherwise, I'm trying to see where it ought to come from.

I just don't know whether the C runtime is supposed to be considered for --no-undefined like Linux does. I doubt any standard specifies that one way or another.

My brain is probably mush today but if I say "extern char *environ[];" and then test "(&environ[0])" it builds.

triaxx: ^

mason: wouldn't have thought the compiler wouldn't have considered these two equal

mason: actually it builds, but o_O

meena: I saw a funny thing noting how arrays and pointers can be seen to be different.

Either way, we document it as **environ so there's a bug somewhere, and I'm going to go out on a limb and say not a docbug.

a limb? I always thought it's going out on a limp

A limb, which gets thinner and more tenuous the further out you go. Limping is after you fall off of it.

Limb as in treelimb.

Title: Dennis Ritchie - Write in C - Invidious

Do you have access to a Linux system? It would be worth going through all the stuff that does get linked in to find the object file or .so file that contains the symbol. Then see whether FreeBSD links a similar file.

i'd like to point out that assembler has less undefined behaviour, and yes, of course pascal won't quite cut it, because the grip of its knife isn't also made of a knife

hehe

C++ fixes all problems with C :-)

cracauer: the fact that you can say that and not get lynched, i think i'm gonna love it here

cracauer: it fixes *some* problems. it can't fix all, because it claims to be C compatible

thephd.dev this person, who works on both, wg14 and wg21, has some really insightful rants

Title: The Pasture | The musings, ideas, discussions, and sometimes silly words from a digital sheep magician

"foo + bar" might not be addition

Which it isn't anyway.

cracauer: bsd.to/MLzJ

Title: dpaste/MLzJ (Plain Text)

cathode.church/@meena/109501287796167199 this quote is from the latest

Title: Mina: "reading: The Wonderfully Terrible World of C and …" - Cathode Church

ok, I failed my copy-paste for Debian: bsd.to/ZhSr

Title: dpaste/ZhSr (Plain Text)

It seems that for both Linux and NetBSD, environ is defined in libc, but not for FreeBSD...

Yeah, and FreeBSD links in libc, too, when making the shared lib.

triaxx: that's really weird, where else would it be?

I noticed on a pkg autoremove that a package asked me to remove a user which had been installed with the package

now I'd like to check if there are any other orphaned users from the past on my system

is there a way to check that automatically?

futunebot: which package? I'm pretty, and sure, the package should do that itself

meena: I think it was trousers and the user was called "_tss"

/usr/lib/crt1.o contains the definition of environ

alternatively, is there a log of the pkg uninstallation message anywhere?

cracauer: Yes, I know

It is linked against when you make an executable. But not when making a shared lib.

I noticed that

I'm diffing the clang -v ld args right now.

So I think a bug report is in order.

cracauer: that is something usually linked, tho

But I think it is a reasonable expectation that environ is defined in the actual libc.

futunebot: yes, it should just be in /var/log/messages

That is why the linker pulls in libc when making a shared lib.

meena: I tried grepping for _tss there and got no hits, maybe I should look manually

I noticed that the linker line still fails on FreeBSD if you add -lc to it.

futunebot: aside from the 80 or so patches, this looks like a normal port: codeberg.org/freebsd/freebsd-ports/src/branch/main/security/trousers do i don't know why it's doing that

Title: freebsd-ports/trousers at main - freebsd-ports - Codeberg.org

futunebot: grep for pkg

meena: I'm getting a hit for trousers deinstalled, but the messages are not preserved there

triaxx: still there?

cracauer: yes

I made up my mind that this is a bug and that the definition of the symbol should be moved to libc.

Which software package first showed the issue?

pulseaudio 16.1

It is 14.2 in ports

Do you want to file a bug?

I am doing

The current definition is in /usr/src/lib/csu/common/ignore_init.c

I would move it to src/lib/libc/stdlib/getenv.c

Title: 268518 – Undefined symbol error when building shared library with environ(7)

Thank you!

Are you officially updating the pulseaudio port?

No, I would like to build a package depending on pulseaudio (using the pkgsrc framework)

I will tackle a bunch of audio ports, mostly jackd related ones. Feel free to coordinate when convenient.

Hi, could anyone help me with IPv6? I have an additional router in my home network, which does RAs for ULAs only. My FreeBSD client (a TrueNAS jail) appears not to be using this at all...

cracauer: Thank you!