checking whether mktime, gmtime, localtime work correctly in 2020... no

configure: error: date-time conversions do not work in 2020 -- is tzdata installed?

Umm, poudriere is complaining about that for math/R. Is that likely me or a port issue?

hm, are you building WITHOUT_ZONEINFO ?

Is that an option for make.conf or src.conf?

src.conf

Nope, don't have WITHOUT_ZONEINFO=yes

is TZ set to anything weird?

Shouldn't be, where am I checking for that? On the host? In the builder jail? Poudriere config?

in the builder jail, I don't know whether poudriere inherits it

Hmm, I wonder

Nope. I'm also assuming that poudriere copies that into the jail after it clones it or whatnot.

do you have WITH_ZONEINFO_LEAPSECONDS_SUPPORT set in src.conf? (never do that)

Nope.

if it's "never do that", why it's even an option?

Oh yeah tz in builder jail is fine. Just looked at the logs and the timestamps match the host.

dunno. I think I put up a doc patch to document the dangers of it

I never spent much effort pushing for it or finding reviewers: reviews.freebsd.org/D25902

Title: ⚙ D25902 Add warning against leapseconds to src.conf manpage.

[00:02:43] Failed ports: math/R:configure

so it's not just you

ah, that's good to know.

have you checked on the package builders?

I should get into the habit of checking bugzilla for existing issues first. Unlikely, but only takes a few seconds to check. Isn't there a CLI tool in ports to help with the search?

there don't seem to have been recent changes to the port

I just started poudriere porttest on it and dropping into a shell to take a lookg at config.log.

what OS version?

1302505 for both kernel on host and userland on host and in jail.

does /usr/share/zoneinfo/Europe/London exist in the builder jail?

Apparently not.

that's odd. it doesn't exist in mine, but that's because I build with WITHOUT_ZONEINFO=yes

here's a workaround:

There's the relevant snippet of the config.log that was dumped by configure.

in make.conf, math_R_VARS+=CONFIGURE_ENV+=r_cv_working_mktime=y (just testing it myself now)

ah

bah, didn't work

I assume that's an R-specific idiom since it's prefixed with r_? (was going to ask if that's a common idiom to force configure check statuses)

oh nevermind then heh

What's exit status 4?

it means something did exit(4)

I meant 4 specifically :P

Was just looking for that but having trouble finding it. Guess maybe it's not standard, even though I just happen to read in the man page that "it's not good practice to call exit with arbitrary codes"

from the chunk of code, it's in the part testing Europe/London for DST

if(res != 1593601200L) exit(4);

oh

hm, try =yes rather than =y

k running now

adding that kind of stuff to CONFIGURE_ENV is indeed a way to force the result of configure tests, but the name of the test varies and you usually have to look them up in ./configure

=yes might have worked, seems to be getting further this time around.

[00:01:05] checking whether mktime, gmtime, localtime work correctly in 2020... (cached) yes

yeah, the (cached) means it's believing the value from the environment var rather than actually doing the test

yup

Should Europe/London exist or is more a case of the configure checking logic being incorrect?

I'm not sure why you don't have the zoneinfo files though. what parameters did you use creating the jail?

Europe/London should exist if the zoneinfo files exist at all, which should be the case unless you built WITHOUT_ZONEINFO=yes

It's from a slightly customized build of stable/13.

(there's a zoneinfo package you can install if the base system is built without zoneinfo, but poudriere can't cope with installing that package, though it can build it)

(since it doesn't like things writing to /usr/share)

I checked src.conf and WITHOUT_ZONEINFO isn't in there, but who knows, maybe the build that I have on my server was built with WITHOUT_ZONEINFO=yes and the line in src.conf was removed after the fact.

I have FreeBSD-zoneinfo since I'm using pkgbase.

I wonder if that's the issue.

what options did you use to poudriere jail -c ?

Hmm, just checked revision history for src.conf in my repo and don't see any sign of WITHOUT_ZONEINFO having been there at any point (which would be odd too because I doubt I would have disabled it).

anyway, the workaround seems to work fine

Don't remember now, but checking /usr/local/etc/poudriere.d/jails/1302505-amd64/* to see if that gives me any hints.

poudriere jail -l should show relevant info

METHOD=tar

Don't tell me the set of tars that I built a long while back is missing those files.

shrug

Feels kind of strange though that an error like this hasn't popped up before if that's the case. I run a bunch of poudriere builds all the time and the jail was built 6+ months ago.

shrug

all I can say is that the R port as it currently stands won't build without zoneinfo files

Yeah, this is going to nag me as it's a bigger problem to me if the dist archives are missing files that should be present :P

Oh well, another item on my TODO list.

meena: Does update-packages require buildkernel+buildworld first or does it include that as part of its target?

you're fun :)

yuripv: changing the on-disk format for PR 265950?

265950 – POSIX 2008 locale failures when global locale not C bugs.freebsd.org/bugzilla/show_bug.cgi?id=265950

nope, adding yet more stupid public symbols to libc :(

ahh

phew :-) I've been working on this kind of stuff for a downstream, and it would've been a weird time in that to drop another format revision :-p

kevans: like this: reviews.freebsd.org/D41927

Title: ⚙ D41927 pr265950 draft

265950 – POSIX 2008 locale failures when global locale not C bugs.freebsd.org/bugzilla/show_bug.cgi?id=265950

yuripv: ah, ok, yeah, that's about what I had pictured... it does suck that we have to export this just to do it right

odd question, is there any easy way to download a file from git hub, without using git? maybe using wget or curl?

sure

or an equivalent tool in base?

something like this

so use the raw.git etch url and use fetch

so fetch raw.githubusercontent.com/someuser/somerepo/master/fileyouwant

yeah, there's generally a 'raw' button for most things, then you can just feed that to fetch(1) whose default behavior is like `curl -O`

(/ `wget`)

awesome, thank you

brilliant, thank you!

_xor: it includes that as part of its target

how can I change vxlan udp/port number on ifconfig commands ?

thanks

man.freebsd.org/cgi/man.cgi?ifconfig(8) vxlanlocalport etc.

Title: ifconfig(8)

Hi Guys

plz, where can I buy valid US CC??

AllanJude, you willing to take the ownership of #eurobsdcon

I have a gentleman trying to prove it should be his, it certainly should not be mine - I took it over during the server move as a defensive move

his emails are not reaching me and he has mentioned you

I trust you so ... yeah

Hi. Looking for a freebsd git merge conflict resolution tool. I see there is linux-bcompare but it requires the compatibility enabled in the kernel. Do i miss another tool?

What makes me look is this git output: 'Hit return to start merge resolution tool (bc): The merge tool bc is not available as 'bcompare''

RhodiumToad: jails could help if you nullfs mount

ooups looks list I was too far in my scroll history

broke something, now su returns su: Sorry

removed user from group wheel?

Add user to dialer group for access to serial port (as root):

 

pw usermod myuser -G dialer

that's what i did

# pw usermod me -G dialer

that's it, yes

but groups return: me wheel operator video vboxusers ?

what would that?

V-T60: did you re-login?

and # groups me gives me dialer

"me dialer"

was that command bad?

presumably i had "me wheel operator video vboxusers"

it overrides the groups for user instead of adding user to group

and now have "me dialer"

i want it to be " me wheel operator video vboxusers dialer"

pw groupmod dialer -m me

how do i reach that result when got into superuser account?

pw groupmod wheel -m me

and so on

okay

-G overrides, i think -m adds

Is there a way to tell which particular driver was used for a device? I read the manual of pcm (snd) and learned a bit about it. I checked and figured out the snd drivers are compiled into my (default) kernel. I checked dmsg and see pcm lines picking up the audio devices. But i dont see which driver was used or do i?

KingShark: try pciconf -l (may be with -v for more information)

also devinfo -v

does it even make sense to use pulseaudio on top of oss? wouldnt oss be enough?

panic: page fault. Is my ram/disk failing? :)

i installed a kernel debugger now in case it happens again

nice

page fault panics aren't usually hardware problems, though they can be (typically RAM rather than disk)

more often they are driver bugs

interesting. did not tinker around much with drivers, just the freebsd 13.2-release base system with the i915kms drivers for intel/nvidia gpus. it happened while using blender so maybe the rendering sent some unusual calls

KingShark: you don't have to change anything in the drivers (or more generally, the kernel) to experience page fault panics. they are usually caused by dereferencing invalid memory (not always, but usually)

it's likely that this is a bug somewhere and if you can get a core dump, you could report it

Installed and configured nginx under FreeBSD. The log files are in /var/log/nginx. Do they automatically get rotated?

dstolfa, it did write something in /var/crash/core.txt.0 but i lacked a kernel debugger at that time so thats the only information in the file (i installed gdb meanwhile)

tercaL: No. I had to add a script and crontab entry to do that

does anyone know why my dns doh nginx gateway is not working? I want to create a DNS server with bind9. `nginx.conf`=pastebin.com/5jCu21CX `named.conf`=pastebin.com/fy0FG1ST

Title: #user nobody;worker_processes 1;events { worker_connections 1024; - Pastebin.com

Schamschula: why crontab though when there's newsyslog(8) and newsyslog.conf(5)?

al1r4d: not working? any error message?

i can't find on /var/log/message.. but i realized i did wrong.. let me fix it

al1r4d: nginx doesn't log to there, usually

since when is the -g option for geli (encrypted /boot, or so it sounds) around? and is there any documentation on how to actually use it? because man geli is pretty tight-lipped about that…

the geli boot options are weird because it's not exactly clear which ones are for gptboot and which for loader

(and which for kernel)

oh, there's a distinction? that's even more confusing because i still configure uefi booted stuff with loader.conf…

and god damn, i downloaded the tiny memstick image to fix up a new system quickly and have to see that the new installer is extremely janky. i think i'll just do it completely manually because figuring out why the damn thing keeps breaking is probably more work :F

like setting up an interface that's already set up, skipping the dhcp step even tho i said yes but then ignoring the existing dns config. failing to download files and then just wanting to completely restart instead of just offering me to choose another mirror. then failing even after restarting the setup because somewhere some partially fetched file is laying around and (surprise!) not

extractable or some other dumb stuff like that…

I slightly revise my previous statement

having picked through various bits of code, the -g flag is used by both gptboot and loader,

while -b is used only by the kernel

gptboot and loader will only try and decrypt partitions where -g was specified

kernel will try and decrypt partitions with either -g or -b

bottom line is that you need -g for partitions containing /boot, and -b for partitions that don't have -g but which contain the root filesystem

for the longest time the GELIBOOT flag didn't imply BOOT, so you could easily create a GELI partition that we'd deceptively use in loader but fail to mount root from it

hm, I didn't check before 13.2

i think i fixed that in 12.0

so use -g for partitions containing /boot and -b for partitions containing / even if they also contain /boot

i just couldn't think of a reason where we would want that

specifying -b on a partition that also has -g does not seem to be harmful

okay, one thing i don't understand tho – where does a -g partition get its keyfile from?

or does that work exclusively for when the passphrase represents the entire key?

the installer was never fixed to use just -b instead of -bg, fwiw

so -bg is actually incredibly well-tested still

gptboot has nowhere to read keyfiles. I don't think that loader has anywhere either, though there is a comment about that feature being missing

grmbl, tiny memstick doesn't have the archives and fetching via ftp.freebsd.org fails…

RhodiumToad: well, i can put a keyfile on /boot and note it down in loader.conf – but that only works *after* you actually got into /boot – hence my confusion.

loader has a load_geli command

(one could in theory have loader.efi get keyfiles from the ESP, but it seems questionable how much security that adds)

what's the ESP?

the efi partition?

ESP = EFI system partition

I think the more common scenario is to load_geli a key from removable media

kevans: how? i currently have a setup where i just have the entire /boot on a usb key and just yank it out after the machine is started.^^

load_geli loads a key, not a keyfile, no?

should be a keyfile according to the manpage

ah, and it sets the module type to geli_keyfile: 267 sprintf(typestr, "%s:geli_keyfile%d", argv[1], num);

right, but presumably loader itself isn't using that, it's for kernel to use?

correct

so you can use that for example to have a keyfile for /, but not for /boot

right, it's for a case like phryk's where they don't care about encrypted kernel

well it's not that i don't care. i just came up with this pattern as the most reasonable compromise back when i did my first freebsd laptop. back then, either -g didn't exist or i overread it in the manuals.

honestly i still want to have a proper secure boot. /boot *could* be unencrypted for all i care as long as it'd be signed and the signature checked on startup so people can't just add a rootkit.

but i think geli with -g might be the next best thing currently available.

I believe you can sign loader.eli, though I haven't looked into whether you can then have that validate /boot files

but if loader.efi is signed and /boot is encrypted, that should be enough for most cases

i currently only have like decade old hardware. i don't think you can load custom keys/sigs into their tpms.

have a new 7950x lying around and gonna get board/mem next month so might be worth a shot with the new equipment.

i also recently stumbled onto something called mfs in the docs and think that could be used to really improve my remote-bootable crypto setup on the live machine.

… what was the command to make the built-in vi not be a complete flustercluck again? i thought ":set nocompatible", but apparently that's an error… :F

mhh, i created an efi partition as second partition on a disk and it doesn't seem to be recognized…

you want vi to be nocompatible with vi? :D

i only created an msdosfs with -F 32 -c 1 and copied loader.efi to EFI/BOOT/BOOTX64.EFI, pretty sure this is enough on my laptop enough…

yuripv: i think the compatibility in that refers to ancient hardware nobody outside a museum still has

it recognizes my laptops bootsticks, too… so i guess this bios just doesn't recognize efi partitions that aren't the first partition on the disk…

phryk: did you set the partition type correctly?

RhodiumToad: pretty sure something-efi

what does gpart show say about it

give me a minute, currently booting into the livestick to set up a custom bootstick^^

RhodiumToad: "efi (40M)"

that should work

though 40M is too small

works flawlessly on my laptop so far.

(bios should still recognize it)

I mean, the recommended size is much larger

why tho? i mean i'm putting one file on it that's… 928K?^^

it's just the standard advice

okay, going the usb stick route seems to work. getting a couple "Can't read linker hints file." but it's running.

weird, freebsd-version seems to look to /boot to figure out the running kernel? instead of just… asking the kernel? that seems a bit backwards

It's using sysctl(8).

debdrup: are you sure? it told me it couldn't determine the running kernel version when i didn't have /boot mounted…

If you look at the shell script in /usr/src/bin/freebsd-version/freebsd-version.sh.in, you'll see that the running-version function uses `sysctl -n kern.osrlease`.

debdrup: mhh, sysctl doesn't fail when /boot is missing, tho. but freebsd-version -k does.

okay, 13.2-p3 working, now updating the fresh system to 14.0-BETA2. here's hoping it doesn't kill another disk the second time around ^^;

finally got a working 14.0 with working drivers for my new gpu.

now working data recovery. wish me luck. :P

Good luck, phryk!

break a leg phryk !

phryk: which GPU? I am running drm-515-kmod for my AMD 6750XT in 14.0

thedaemon: rx6600xt

and thanks, nothing exploded yet, currently backing up all development data ^-^

yay

Freaky: any idea how to integrate this, freshbsd.org/freebsd/src/commit/da5…0f586389de0063650343e264ae58116e3c0 so that commits like freshbsd.org/freebsd/src/commit/948…1aaf420cd7d493cc2e118cacc06d18653fe and freshbsd.org/freebsd/src/commit/1c1…7979571bf07c05a48e857b7b285b037410f show up when searching for me

Title: FreeBSD / src / da51e0f / Add .mailmap to normalize some author data - FreshBSD

what tool is that .mailmap for?

yuripv: i think it's for git

ah right, thanks

looks like cgit doesn't care either cgit.freebsd.org/src/commit/?id=948…1aaf420cd7d493cc2e118cacc06d18653fe (yet)

Title: src - FreeBSD source tree

looks correct to me?

or if you mean 'c' in your name, looks like it's missing the "Mina Galić <freebsd⊙ic> <freebsd⊙ic>" entry

yuripv: wanna comment on the review?

I didn't know that's necessary

i was just happy is possible

yep, done

yuripv: oh, for entries that already have the correct email address

correct

yuripv: you want to do the honours?

i can do it, yes :D

meena: that didn't help cgit, but it works in git log now

yuripv: cool

thank you!

meena: hmm, I'm not sure, I'll add it to the TODO

hi. how can I know what installed /usr/bin/tar on FreeBSD? I would like to get the source code of tar. "pkg which /usr/bin/tar" tells me that it's not pkg. i installed ports, but "find /usr/ports -type d -name tar" doesn't find anything

/usr/bin is FreeBSD

i searched for "tar" in github.com/freebsd/freebsd-src but hum, they are many words which contain "tar", and so it's a pain to navigate the results :-)

Title: GitHub - freebsd/freebsd-src: The FreeBSD src tree publish-only repository. Experimenting with 'simple' pull requests....

Title: wtf · GitHub

no.

VimDiesel: if you look at usr.bin/tar you'll see the Makefile refers to contrib/libarchive

it's usr.bin/tar, but that's only Makefiles, the real source is in contrib/libarchive

thedaemon: is it github.com/freebsd/freebsd-src/tree/main/usr.bin/tar ? this directory doesn't contain C files

Title: freebsd-src/usr.bin/tar at main · freebsd/freebsd-src · GitHub

er, vstinner:

yuripv: latest tig uses it correctly. So next up, bug report for cgit

/usr/src/contrib/libarchive/tar

It's here ^

if you have source downloaded

Title: Chapter 26. Updating and Upgrading FreeBSD | FreeBSD Documentation Portal

meena: latest what?

Title: GitHub - jonas/tig: Text-mode interface for git

oh

it's my preferred way of reading commits when I'm on my computer. (when I'm on my phone it's freshbsd ;)

'git pull /usr/src' this should get you going

oh neat meena

im running retroarch on the console with the i915 drm driver and i cant figure out how to give a user input permissions

I can't believe I've only been contributing since 2021. it feels like an eternity.

theres no input group

so input only works if i run it as root, but not as a regular user

what groups is your root in that user isn't?

groups rarely matter for root

thedaemon: on standard Unix, root doesn't really have to be in any special groups

oh my bad

but for drm stuff, the user should be in the video group

root has to be in video group for x11

in FreeBSD at least

:D

i have the user in the video group and retroarch runs and video works

but i cant figure out how to make the keyboard work

thedaemon, yuripv : ok thanks, i found what i needed :-) i'm fixing test_tarfile of Python on FreeBSD: python/cpython #108950#issuecomment-1730361505 -- in short, FreeBSD clears the sticky bit, unless -p option is used

Title: gh-108948: tarfile should handle sticky bit in FreeBSD by sorcio · Pull Request #108950 · python/cpython · GitHub

on linux you just add the user to the input group but on freebsd there is no input group

is there a man page?

I don't have it installed

or check the port doc, did it have anything when you installed?

bjornn: what kind of input are we talking about here? joysticks?

no

what isn't working? keyboard input?

yeah the keyboard doesnt work

ok

just making sure lol

it uses libudev-devd i think

Title: FreshPorts -- games/retroarch: Cross-platform entertainment system based on libretro API

which like makes freebsd's devd work with apps that use linux udev

you can also use webcamd

you have to grant access permissions on the appropriate

/dev/input/eventN device to your user account

^ found this as the pkg-message

and this is how the pkg was built 'UDEV=on: udev device hotplug support'

may be we could have input group as well and make those root:input 660

i suppose i could just use the operator group

and make /dev/input like the other dev folders that use operator

or just create an input group

but i assumed devd would already have some functionality for doing this

ok i think i figured it out

you dont manually mess with the permissions of anything in dev

you use devfs.rules

this really ought to be documented

but i suppose its not an issue for anyone uses xorg

good to know

you running in FB or something?

or wayland?

fb

ok yeah heres the way to do it

[localrules=10]

add path 'input/*' mode 0660 group input

this has to go in /etc/devfs.rules

and then you add devfs_system_ruleset="localrules" to /etc/rc.conf

and then create the input group and add your user to it

WARNING: Kernel has no file descriptor comparison support: No such file or directory

im also getting this warning when i start retroarch

not sure what that is about

im starting to think it might be easier to just use xorg for this

the drm stuff is all so linux specific

i miss when free software was actually written to be crossplatform and linux wasn't the sole OS that anyone cared about supporting

that was when? :)

like 20 years ago

we already had drm ported from linux back then

bugs.php.net/bug.php?id=77284 oh

Title: PHP :: Bug #77284 :: base64 broken on FreeBSD

77284 – Update port: databases/py-MySQLdb new default options bugs.freebsd.org/bugzilla/show_bug.cgi?id=77284

Title: Disable ifuncs on FreeBSD · php/php-src@2915891 · GitHub

I'm currently playing with ifuncs and trying to see how to integrate it with Ruby, encountered that - PHP has them disabled for FreeBSD

why?

Title: php-src/configure.ac at master · php/php-src · GitHub

meena: why to which bit of that

i really hope whoever is maintaing the port has a patch to enable it for recent versions of FreeBSD

doesn't seem to

we are making use of ifuncs in base / kernel? csu? etc… so i bet we'd notice if it didn't work.

php/php-src 5a361c3it seems the underlying problem was their end

FUZxxl: ^ your code uses ifuncs, right? wanna chime in on that php issue

Freaky: i don't wanna look at php code ever again

anyway, it's bed time, i just don't wanna get up. otoh, sleeping in this chair will not be comfy

this chair bi-phobic!

meena: libc certainly appears to use them

I should dig out of old SSE2 memchr port

s/of/my/

I'm doing byte counts at the moment, I don't think there's a libc function for that though

be good for wc -l

cool

yuripv: git.zx2c4.com/cgit/about there's no issues. just mailing lists. do these people not know that I hate mailing lists?

Title: cgit - A hyperfast web frontend for git repositories written in C.

"hyperfast" wonder if there's marks of benches

meena: one day I want to make my own issue tracker, for a laugh

LxGHTNxNG: can't be worse than no issue tracker

LxGHTNxNG: feature request: i would like to be able to search existing issues, and quickly figure out if they are already fixed

I should easily be able to bolt this on the side of a mailing list