hi

hey

on freebsd 15 current and gnome havn`t access to settings

on terminal gnome- tab can find settings

tweaks yer but no settings

i would like do an audio and video workstation developpement but the configuration don't go

how can i repport bug ?

buzilla

ok polyex i submit

wish i could help more buddy

polyex: how did your DNS filter solution went?

haven't done it yet. trying to get dns working during a scripted bsdinstall right now :(

how bout you kenrap?

I'm taking a break from it for now

tempted to just got with pi-hole or adguard home, but I like the feeling of having more control while doing it in FreeBSD

*go with

kenrap : unbound+aguard home

during a scripted bsdinstall, before the reboot, how can i get dns resolution to work? service netif start brings up IP and i can ping my router (but no public IPs for some reason) but can't resolve domains. i added "nameserver x.x.x.x" to /etc/resolv.conf, and tried service resolv start. nothing works but i don't wanna use dhcp if there's a way

Check if local_unbound is running.

i'm sure it's not because nothing in my bsdinstall script enables or starts it

i guess i'm assuming but how could it be?

the files it creates on start up, /etc/unbound/* don't exist after the reboot so i'm sure it hasn't run

You could drop down to prompt and do a ps aux | grep unbound

??

sure sec

Is there a reason you want to do that before restarting?

ya no processes found

well because i want to make it so i can start a new box with a custom installer and once it reboots it's 100% ready to go

so that means fully setting it up during the scripted bsdinstall

before first reboot

Well, are you modifying the /etc/resolv.conf on the ramdisk or on the mounted partition?

i can get dns working if i use dhcp but i'd rather use the static ip config i already have in place

not sure

It's been a while since I installed FreeBSD. What does df -h show?

in bsdinstall script i just modify whatever's at /etc/resolv.conf

sec

ok i got the output

al1r4d: could you rephrase what you mean more specifically?

i had to record it on my phone tho BillyJoeBob so can i just tell you the info you want from it?

How many lines is it? You can paste it here: paste.debian.net

Title: Debian Pastezone

like 15

the /dev/ufs/freebsd_installer then a devfs, tmpfs, bunch of zroots, a /dev/gpt/efiboot, another devfs

How big is the drive you're installing it to? Find the same entry that has about the same capacity space as the drive you're installing to.

20GB

that's the zroot

Where is that mounted to?

oh hm, /mnt

Like under /mnt?

ya

do a "more /mnt/etc/resolv.conf"

no such file or dir

ls /mnt

no output from that

Yeah, I don't know what to tell you. It's been too long for me since I've installed FreeBSD and I've never dealt with scripting on the installer.

ok np tyvm

RhodiumToad pls ping me when around maybe you know

Worst case, reboot and see if it's correct. Or just start the process in a VM and keep doing it until it works lol

polyex i thought you want to build dns server 🙂

what?

yes, dns server

no he wanted a dns proxy...

but if not, just forget it, sorry for false thought 😅

i have 2 separate needs guys

What? polyex

how do i direct stdin to a command?

I think it's (for example) command < file

no such file or dir

oh wait

i need to pass some text to resolvconf -a vtnet0 through stdin

not sure how

oh just the echo pipe trick nvm

ok i still can't get dns resolving in a scripted bsdinstall env

damn

during a scripted bsdinstall, before the reboot, how can i get dns resolution to work without using local_unbound? service netif start brings up IP and i can ping my lan/router (but no public IPs for some reason) but can't resolve domains. i added "nameserver x.x.x.x" to /etc/resolv.conf, and tried service resolv start. nothing works but i don't

wanna use dhcp if there's a way

uh

do you really need to run netstart

netstart?

that's not still used i don't think

add a sleep maybe?

it really should happen right away

"it"?

the resolver function

add a sleep where?

forget it

hmm

there's no defaultrouter

that would do it

hm why can't i get defaultrouter to show up in netstat -rn?

i wouldn't use netif; i'd set the interface, add the route, set the resolv and you should be good

omg got it

is there a way to add a defaultrouter transiently? so that i don't have to add it to /etc/rc.conf

"route add default 10.11.77.1"

wow that seems to work!!

ok this is weird. in the new system i can resolve www.openbsd.org but not www.freebsd.org lol

can resolve netbsd.org too wtf

o rly

are you using a local resolver or a public one

my isp

try www.geo.freebsd.org

instead

this is weird too. when dns does eventually resolve for www.freebsd.org, i can ping www.openbsd.org but www.freebsd.org fails no route to host

weird ipv6 ?

www.geo.freebsd.org unknown host

dunno

btw is there a way to transiently set a name server for a host like route add default x.x.x.x for defaultrouter?

without having to add something to /etc/resolv.conf

i don't think that there is.

what's the lightest way to set a dns server so that host foo.com will resolve? is it echo 'nameserver x.x.x.x' | resolvconf -a vtnet0, or just echo "nameserver x.x.x.x" > /etc/resolv.conf, or?

i'd just pipe it, as i said i haven't really used resolvconf

when unbound gets installed it doesn't reconfigure /etc/resolv.conf. shouldn't it do that so it works?

like change the nameserver to localhost

i have this in my make.conf DEFAULT_VERSIONS+=perl5=5.32 and trying to upgrade znc and it wants to upgrade my perl to 5.34 instead

n30: perl 5.32 just went EOL

so i cant keep it at all? have other packages i cant upgrade yet due to other issues so i need to stick with 5.32

polyex: did you enable and start the unbound service?

n30: you can keep it, but then you have to fork the ports tree

meena so i gotta jump through some hoops

the hoops are: you have to configure a service you want to use

meena i boot scripted bsdinstall, script adds nameserver to resolvconf, restarts netif, adds defaultrouter, bootstraps pkg, installs unbound pkg, THEN needs to convert /etc/resolv.conf to being compatible with unbound (nameserver 127.0.0.1). not sure how to do the last step yet

no i mean i gotta set up temporary dns, then change it

doesn’t local_unbound at least have a script to do that?

if it does, it's not documented anywhere

no i'm using unbound not local_unbound

local_unbound is never enabled or started

oh, cool, I'm the last person who who's commit is in contrib/unbound

in csh i can redirect a string to a command over stdin like 'echo string | command' but how do i do it if i want to redirect multiple lines into the command?

echo -e "string 1\nstring 2"

polyex: why are you scripting in csh?

csh is decidedly not for scripting

omg i got it

i got scripted bsdinstall to bring up network and dns so i could install packages. no dhcp needed

then after it runs, bsdinstall script overwrites /etc/resolv.conf with the permanent settings (nameserver 127.0.0.1)

there any benefit to using resolvconf -a ... to set /etc/resolv.conf rather than just writing the file directly?

🤷🏻‍♀️

polyex: congrats on getting bsdinstall scripted

ty!!!

meena: What kinda emoji is that?

Some kinda combination? shrug, something, female?

skered know any benefit to using resolvconf instead of writing /etc/resolv.conf directly?

skered: the something is my skin colour. I haven't seen the sun since i moved away from Vienna

Ok so a shruging female is shrug, 200d, female...

Makde kitty, tmux, epic lose its mind... multi multibyte emojis

Made*

skered: yeah, I'm not sure there's any terminal that can display that properly

makes me wonder about other Unicode ranges that heavily rely on ZWJ, like Perso-Arabic Script

سَلَام‎

should probably still display, but who knows?

those little ticks over letters are (probably) vowels

I'm on 14 BETA3 and I think I got a memory leak somewhere. things (firefox, mpv, …) keep getting oomkilled. also whenever there's heavy disk io, my system pretty much completely freezes up…

any idea on how to triage this? i keep looking at top. judging by res there should be no memory exhaustion, but judging by size, at least firefox allocates crazy amounts… like multiple ff processes that allocate more memory than the entire system has.

what do the ram/swap lines look like?

a lot of FF's memory is shared between processes, which is why it doesn't add up to a reasonable amount

no swap

tmpfs /tmp tmpfs rw,mode=777 0 2

how much ram?

12G

what does the Mem: line in top look like?

and how much space is that tmpfs using?

Mem: 3852K Active, 53M Inact, 18M Laundry, 10G Wired, 1145M Buf, 257M Free

10G wired... using zfs?

252M

no, ufs only.

but with geli.

what are you running that wires memory?

bhyve?

nope. nothing much. currently not even firefox is running.

x11, awesome, half a dozen kitty terminals, claws-mail, ncmpcpp, dino, weechat and that's currently it.

all of that should be pretty slim.

can you paste your vmstat -m and vmstat -z to a paste site?

sure, sec

uuh

RhodiumToad: i wanted to paste, but when i started firefox, both displays just… turned off. had to reboot. when shutting down, i did get the kms vt back, tho…

polyex: you are still on about a temporary resolv.conf

RhodiumToad: vmstat -m: paste.xinu.at/mSN vmstat -z: paste.xinu.at/S9S

Title: vmstatm

(captured before my displays went awol)

in future you may find cat ... | nc termbin.com 9999 useful

or use curl to paste to dpaste.org

yeah, or sprunge, but whenever i actually need that i never quite remember the right call :P

firefox in general is also super sluggish. like it taking multiple seconds to scroll, focus/type into an input etc, tabs constantly crashing. in the beginning i thought it was firefox disk cache because i first noticed it on heavy io load, which is when this behavior shows the most.

rtprio ya you got something?

what drm driver are you using and was it compiled from ports or installed as a package?

hey RhodiumToad there any benefit to using resolvconf -a ... to set /etc/resolv.conf rather than just writing the file directly?

installed as package. drm-kmod-515. it's essentially the entire reason i went to beta – to finally have support for my rx6600xt

polyex: using resolvconf means that what happens is configurable via resolvconf.conf

phryk: a bunch of this excessive memory usage is tied to the drm driver. so something is not happy there

from the vmstat -m:

lkpikmalloc 6103476 6081121K 199208952 16,32,64,128,256,384,512,1024,2048,4096,8192,16384,32768,65536

that's the linux kpi, which is mainly there to support the drm driver

i also installed some metapackage to get all the firmware packages. previously only had the one specific to my gpu (dimgrey cavefish) but with that disconnecting one display would cause the other to blank. fixing the metapackage fixed that.

from vmstat -z:

drm_sched_fence: 184, 0, 6068053, 590,21244699, 0, 0, 0

and

malloc-1024: 1024, 0, 6069199, 177,44739201, 0, 0, 0

so… the driver for my gpu on drm 5.15 is likely just broken?

that malloc-1024 is probably what's backing the lkpi mallocs

it might be specific to your gpu or it might be a more general issue with drm

RhodiumToad is that generally a good idea? to funnel resolv.conf changes through resolvconf

polyex: tons of stuff wants to write to resolv.conf, so yes

ty!!

RhodiumToad: so what's my course of action, except for reporting the leak to drm-kmod maintainers?

phryk: I'd try compiling the drm kmods yourself, and if that doesn't help, report it as a bug (or report it anyway)

RhodiumToad: any particular branch of the ports repo i want for 14, or just main?

main

aye, thanks. checking out now.

the ports repo isn't specific to fbsd version, it just has main and quarterly branches

that was my understanding, but i never ran beta before, so better to ask once instead of regretting twice ^^;

freebsd 14 have rust in the kernel?

it makes better code so freebsd better catch up to linux bringing lots of rust in

Rust in the kernel is a complex matter. slightly more complex than, bring it in

"it makes better code". Citation needed. :P

As much as I like Rust, I don't think it matters a whole lot. They're both just different systems programming languages.

With similar performance.

the day rust shows up in core is the day I fork UnoxidizedBSD unless someone else beats me to it

RhodiumToad: I mean, I don't think it's THAT big of a deal. I just don't think it needs to be done.

polyex: I mean, I'm sure you can go ahead and make kernel modules just fine in rust if you wanted to.

they left already

BillyJoeBob: the benefits that rust brings to the table wrt to "simple" programming issues like off by ones and integer overflows are only outshined by the entire class of memory related bugs are hard to replicate in any other systems programming language.

however, many of the things people commonly associate with the downside of rust, as well as many of rust's strengths stem from its standard library and many thousands libraries in cargo

RhodiumToad: i need the os sources to compile drm-515-kmod – what branch do i use? the releng or the stable one?

there's every chance that the standard library is unusable in kernel code, and with it roughly 90% of cargo

everyone will just write in rust in c (using unsafe or whatever it is called)

nah, that's not what I'm saying

just like old joke about fortran programmers writing in fortran in any language

phryk: releng

RhodiumToad: tyvm

meena: i'm not commenting on what you are saying, rather on "better code" above

what im saying is more nuanced. you can get a lot of milage out of rust as Systems programming language without using unsafe. I'm fact, using unsafe, because it's so obvious, is highly scrutinised in the ecosystem and the community. but: you can't get much mileage out of rust ecosystem for kernel development. at least not until someone leads the

way, creating useful no_std libraries. would be a shame if they were all GPL licensed, tho…

AGPL the network stack. >:3

(scnr)

built drm-515-kmod locally and installed it. rebooting now, wish me luck^^

🤞

nothing exploded, now gonna have a look at the lkpikmalloc thing while i do some memory/gpu intensive browser foo. :)

has anyone used any zpool.d scripts?

the ones in openzfs seem perhaps too linux-y and don't seem to work

phryk: visiting any website made after 2006?

meena: yes. :P

RhodiumToad: do you have a ballpark value for a reasonable upper bound for lkpikmalloc? system feels responsive, but the value has been climbing pretty steadily. currently at ~150Mbyte.

okay, it just keeps growing and growing.

how do i properly build the gpu-firmware-amd-kmod port? it doesn't have any options but building and installing it yields only the aldebaran firmware and none of the others…

mhh, they're flavors. i know poudriere can do <origin>@all to build all flavors, but i'm not aware how to do this with raw ports + make

phryk: make FLAVOR=whatever but I don't think you can do 'all' like that

make -V FLAVORS will show the list of available flavors for a port

if the value keeps growing then that's a problem

RhodiumToad: yeah, thought so, especially after it didn't even stop growing after quitting firefox. so now i want to retry with firmwares from current ports.

nope, still leaking

guess i'll report this^^

i am very excited for this bug report, phryk

meena: you are? i just posted it: freebsd/drm-kmod #258

Title: drm-515-kmod amdgpu memory leak · Issue #258 · freebsd/drm-kmod · GitHub

258 – The dialog_menu function's scroll saving feature is broken. bugs.freebsd.org/bugzilla/show_bug.cgi?id=258

and thanks a bunch for helping triage this RhodiumToad :)

huh, VimDiesel talks to itself.

bots do that

anyhow, thanks for all the help, i'll go boot the bad os and play some games. :P

I was gonna play Zelda today, but then i just did house work and napped.

i love naps

Why does this work: "git clone ssh://i4industry-git/v1/repos/insight --progress 2>&1 | stdbuf --output=L tr '\r' '\n' | perl -ne 'print "$1\n" if /(\d{,3})%/'"

< but not "git clone ssh://i4industry-git/v1/repos/insight --progress 2>&1 | stdbuf --output=L tr '\r' '\n' | perl -ne 'print "$1\n" if /(\d{,3})%/' | xargs echo "

<What buffering problem is this

Why does this work: "git clone ssh://i4industry-git/v1/repos/insight --progress 2>&1 | stdbuf --output=L tr '\r' '\n' | perl -ne 'print "$1\n" if /(\d{,3})%/'"

but not "git clone ssh://i4industry-git/v1/repos/insight --progress 2>&1 | stdbuf --output=L tr '\r' '\n' | perl -ne 'print "$1\n" if /(\d{,3})%/' | xargs echo ". What buffering problem is this

Please clarify what you mean when you say works and does not work. That's rather a detailed command line to know what is meant just by inspection.

For example I wonder... Why translate CR to NL? Why not simply discard CR characters?

For example I wonder... The perl won't be line buffered. So won't things buffer there?

why are you piping git clone to perl

I can see that in the perl it is attempting to extract the percentage complete numbers. 30%, 40%, 50%, and so on, with the \d{,3}% pattern.

But I can't guess at why that result is being piped to "xargs echo".

perl will buffer output if it's not to a tty, like most programs

rwp, xargs echo is just a placeholder

I pipe it to dialog --gauge to show a status bar

I should not have spent any time on it but I did. I see that the translation from CR to NL is because git clone is overwriting the line only with CRs and no NLs at all.

translating CR to NL makes it so that each new percantage update (which is being printed with CR <some-number>%) gets translated into a new line, thereby working with the diamond operator in perl

The perl is buffering as we all noted. The $| var can be set to 1 to cause it to be line buffered.

The xargs echo is also aggregating input. Use -L1 to avoid that.

Try this command: git clone --progress ssh://i4industry-git/v1/repos/insight 2>&1 | stdbuf --output=L tr '\r' '\n' | perl -lne '$| = 1; print $1 if /(\d{1,3})%/' | xargs -L1 echo

Also the perl regular expression {, was invalid on my installed versions of perl. Had to convert that to {1, in order to have it work.

the $| = 1 worked

\o/

I tried using it, but ignored the -L1 on the xargs..

ahh god

I wouldn't quite go that far. But you are welcome.

Thank you very much and sorry for making it hard (I was in the show :)

shower*

Glad to hear everything is all cleaned up now. Hahaha.

also

what is the -l flag for

I am on alpine... so its not the same

The perl -l option strips newlines on reading and appends newlines on writing. It's one of the helper options to make one-liners easier.

Also... Let me gently nudge you to keep topics on the right topic channel. Since you were not even close to FreeBSD while having this problem. It wasn't even in the room!

meena: I'm back in town and back on my regular schedule. Let me know if you've found anything regarding vultr credits or whatnot. Can look at taking care of it this week.

not sure if it's a beta thing or my misunderstanding, but I'm on BETA4 right now. if i "bectl list" I see a BETA3 item as "NR". but i'm clearly in BETA4 kernel.

should the BETA4 item be NR?

Title: dpaste/EN03 (Plain Text)

freebsd-version -ku ?

14.0-BETA4 (shows twice on two lines)

I've never explicitly created any boot environments myself nor used anything but the default one

I always use "freebsd-version -kru" to show all three things. -k for the installed kernel and -r for the running kernel. Reboot if those two are different. -u for the userland.

How did you upgrade to BETA4? I think that's where things diverged.

-kru all 3 are the same BETA4

Oh wow, I've alwayes assumed -k shows the running kernel

I upgraded with freebsd-update

Just yesterday I did 13.2 -> BETA3 -> BETA4 upgrade with freebsd-update, everything is fine for me atleast

Hmm... I have always seen freebsd-update create a new boot environment itself okay and all.

rwp: it creates them, it's just not switching mine

i did break my system months ago and used boot environments to revert

i wonder if something needs to be set to allow it to not be static?

For me when I am "cruising along" through the upgrades I am riding "default". You don't have one. So maybe that is where things diverged?

what's weird though, is why is the BETA boot environment booting BETA4?

*BETA3

Full disclosure: I don't know. I use BEs and they are awesome but I don't have every detail in my head about them.

seems like I should be booting into BETA3 as long as that item is "NR"

I think freebsd-update is designed to work with "default". I think if that was deleted then maybe it won't do this automatically anymore.

dh, Can you pastebin your beadm list? Or rather I should ask if you are on "default" too? Or a named one?

I'm on default

That seems to reinforce that not having a default is likely why things are diverged.

Title: EN03

I have no default

Is that required?

Title: dpaste/sruN (Plain Text)

I think "default" is required for freebsd-update to function on it automatically. I think. Not sure. Seems plausible.

And markmcb my request for the pastebin was for dh since the upgrade was normal there. Thanks though but that was the same paste you had posted before. :)

And I am not myself running BETA so mine would not be useful.

I guess I'll try renaming the lastest to default and see what happens on BETA5/RC1

I think that reconfiguring things so that you are back on "default" should fix the automation for the next upgrade.

And right now I think you are booted to 14.0-BETA3_2023-09-23_205902 and that on that clone branch your system is BETA4.

ok, renamed and activated, going to reboot to default then will see what happens next weekend :)

markmcb, What does "df -h /var/run" report for your system? I ask because then I would ask you to "cat /etc/os-release" and verify that it is BETA4 there.

zroot/ROOT/14.0-BETA3_2023-09-23_205902

Normally /etc/os-release is a symlink to ../var/run/os-release

All of that is making sense as to the way things got into this state. And I think you have the right fix.

Just fyi on my system df -h /var/run reports zroot/ROOT/default

Missed him by "that" much. lol

This is reminding me that I have some old boot environments just holding space from 12.x days that I will never need again. I should prune them.

markmcb, Just fyi on my system df -h /var/run reports zroot/ROOT/default which might be considered normal.

Not that doing other things is a problem. These are all tools in the toolbox to be used as required. :-)

made the switch and rebooted. also seeing zroot/ROOT/default now

I am thinking that should automatically do everything for the next BETA5 upgrade. Seems plausible.

If you really wanted to double-blind test things you could revert your system back to BETA3, reconfigure to default, then upgrade to BETA4 again and verify. :-)

Seems like a lot of work at this point though. So not really suggesting it.

yeah, i'll wait. lol.

fwiw, i had a "default" when i first noticed this behavior

so i'm not sure this will resolve it

but we'll see

It's still beta. It might be actual breakage. But dh did report it worked there. So, no idea!

yeah, seems odd behavior. i'd think on an upgrade no matter what the existing boot environment setup, the "R" flag should be set to the newly created environment.

I think it is the opposite. The newly created boot environment is the saved off old one. The currently running one is upgraded. It leaves a trail of previous BEs behind. Pretty sure.

That's not to say that one can't do it other ways such as the way you mention. Just that it isn't the way freebsd-update is coded to work.

hello

can`t acces to settings on gnome on 15 current, anyone can help me ?

need dvorak intl with death key for euro symbol

and switch right clic te left clic

anyone using 14.0 can test TCC? TinyC Compiler? It's giving me a seg fault on a hello world.

thedaemon: I only got 15.0-CURRENT in close reach.

you got TCC meena ?

thedaemon: core dump!

I got mine set on auto but haven't seen it make any dumps

thedaemon: not the most informative coredump

I might have disabled it someother way

oh you mean you got a core dump?

* thread #1, name = 'tcc', stop reason = signal SIGSEGV

* frame #0: 0x0000000000233f66

frame #1: 0x000007cf7cd4421e

let's rebuild WITH_DEBUG and see what that gives us

perl5>=5.32.r0<5.33 : lang/perl5.32 hrmpfvzfsddd...

WITH_DEBUG=yes did … nothing

oh crap

well, I am not smart enough to debug it yet.

Just diving into C.

I'm suspecting this might be related to the libc / csu changes

what is csu?

C start up

yeah I think I was able to get a lib error once

is it tcc itself dropping core?

RhodiumToad: yes

the above lldb output is from that

even tho it's compiled with debug symbols, I'm not getting anything better than that

and the numbers are actually the same.

rearranging my room, bbl.

setxkbmap dvorak-intl

Error loading new keyboard description

lemme see what 13.x says

well, this is way better:

In file included from hello.c:1:

In file included from /usr/include/stdio.h:43:

/usr/include/sys/_types.h:107: error: ',' expected (got "__aligned")

at least it's not segfaulting.

and it's not simd related: fish: Job 1, 'env ARCHLEVEL=scalar tcc hello.c' terminated by signal SIGSEGV (Address boundary error)

(probably)

(but we can still probably ask FUZxxl to take a peek)

what is the disassembly around the fault?

haven't look yet

wait, how do i do that.

error: core file does not contain 0x233f66

error: Failed to disassemble memory at 0x00233f66.

this core file won't let you dis it

oh, so it probably jumped into nowhere

what's in the registers?

sec, bringing the dog out. bbiab

RhodiumToad: gist.github.com/igalic/012bbbfb389dcadbc565d2b3ca20e0eb

Title: gist:012bbbfb389dcadbc565d2b3ca20e0eb · GitHub

thedaemon: ^ also for you, more or less

given that they're mostly 0, we can probably ignore the FP registers

r8 and r9 look funny, but I'm not sure they look suspicious.

Frame #0 has the same data as `rip`

i wonder how painful it would be to run tcc in the debugger…

oh, great, I have to run lldb as root…

In file included from /usr/include/stdio.h:42:

Process 33037 exited with status = 1 (0x00000001)

/usr/include/sys/_types.h:58: error: #error unsupported long size

(no, no it's not. i invoked it with -c hello.c, not with -o hello hello.c)

Title: gist:dab6bde4f06c85406a6a2c2476e194ec · GitHub

that's a tad more useful!

anyway, time to goto bed

hot take: debug symbols should be the default.

go and build rustc with debug symbols on, and notice how much disk space it took

RhodiumToad: i feel like disk space is cheaper than the time it takes n people to build $software just to get the symbols.

tcc with debug symbols is a whopping 240K

somebody should tell FUZxxl about gist.github.com/igalic/dab6bde4f06c85406a6a2c2476e194ec i feel like he might know what's up, or who to bug about this

Title: gist:dab6bde4f06c85406a6a2c2476e194ec · GitHub

it's a null pointer dereference

what's at that line in tcc_load_object_file ?

RhodiumToad: how do I figure that out? oh, by looking at the code.

lldb should be able to show you the source if you tell it where to find it.

or, I could just cd into the /usr/ports…work dir

or… not…

why do i get into these terrible debugging sesssions, that I have nothing to gain from… when it's bed time? always?

Sorry :(

thedaemon: your turn now ;)

cd /usr/ports/lang/tcc ; git pull ; make install-missing-packages ; make WITH_DEBUG=YES BATCH=YES ; make reinstall

that last step isn't strictly necessary, as you can just cd work/$tcc_src and lldb from there