I haven't used Vultr, but I'd start by looking at /boot/loader.conf and /boot/device.hints.

For both the off-the-shelf install and their customized ISO.

Has anyone experienced issues logging into certain sites with firefox? I keep getting browser not supported, I don't have this problem on Debian though

V_PauAmma_V: Ok! Thanks :)

I already copied loader.conf, but did not check device.hints

V_PauAmma_V: device.hints are identical

are they both using the generic kernel?

yes, same kernel

there is one difference. The "fast" is saying Hypervisor: Origin = "KVMKVMKVM" whereas the slow is saying Hypervisor: Origin = "Microsoft Hv"

The problem is the Hypervisor

so the provider is using one hypervisor for its version of the config, but a different one if you do a manual install?

RhodiumToad: exactly

RhodiumToad: when I create the server using custom ISO, it starts in Hyper-V. However, if I ask to use FreeBSD, it uses KVM.

RhodiumToad: solution: I will start a FreeBSD server, and format everything :D

worked perfectly :)

fancycade: it's probably related to user-agent string not having Windows/Linux/macOS? try one of the addons that fakes the reply

some sites I've seen return a 403 error if they see the string "amd64" in the user-agent.

(and yes, I noticed that, ms teams (IIRC) didn't want to work with default string on FreeBSD)

what's your opinion on FreeBSD hosting operators support these days? I've tried vultr and linode, both required manual intervention when it comes to VM creation, which interrupts with automatic deployments, unfortunately

I've written pretty extensive ansible roles for my own #selfhosted stuff on FreeBSD, needed to switch to Linux because of that

RhodiumToad: Hmm, interesting. So I commented out the WITHOUT_DEBUG_FILES=yes and WITHOUT_KERNEL_SYMBOLS=yes lines and re-ran make clean buildworld

RhodiumToad: It hasn't bombed yet. Not sure if it has gotten to lib32 yet or not, but I don't remember it getting this far last time when it did bomb.

Ronis_BR: boot verbose perhaps to see where it hangs

Ronis_BR: it could be something stupid like choosing a timecounter

Is there an easy place to check where boot-time related configuration should go? (e.g. /boot/loader.conf tunables + *_load modules vs /etc/rc.conf:kld_list + /etc/sysctl.conf, things like that)

Last I looked it a bit all over the place, where some things were required in /boot/loader.conf (read-only tunables that needed to be set early?) vs preference for /etc/rc.conf or /etc/sysctl.conf, etc.

sysctl -T and -W can be used to detect what needs to be set in loader.conf

Oh nice, didn't know about those flags.

(if a sysctl shows up in -T but not in -W, it can only be set in loader.conf)

Yup, that's I figured reading the descriptions.

in a shell with process substitution, comm -23 <(sysctl -T -a -N | sort) <(sysctl -W -a -N | sort) gets the list

Oh that's a nice little shortcut. Jotting that one down.

for modules, kld_list is preferred especially for modules of any size

Is there anything that specifically has to go in loader.conf and cazn't go in kld_list?

(on EFI in particular, it's possible to run out of memory in the loader by loading large modules)

anything you need to find the root filesystem needs to be in loader.conf

ah

but then there's a problem: kld_list is processed at a fairly arbitrary time during boot

What about zfs? I know gptzfsloader & loader.efi can both find zpools, but do they also implement enough reading capability for use during boot? (just visualizing & guessing here)

for most rc.d services, they know how to load their own modules, but there are cases like the file system firewall (ugidfw) where kld_list is just too random

...or does zfs_load still need to be in loader.conf?

zfs_load needs to be in loader.conf in order for the kernel to be able to mount zfs roots

loader contains its own copy of zfs to be able to read /boot, but that doesn't help the kernel

Oh right

Makes sense.

Hmm, from the looks of it, a vast majority of these lines should be movable to /etc/rc.conf and /etc/sysctl.conf.

what do you have there?

oof

I should clean that up, don't need some of those lines. Though I'm going to keep the ones with the default values set explicitly (they're for future reference in case I forget).

I remember tmpfs_load causing a couple of boot-time errors about how it's already loaded.

tmpfs is usually compiled into the kernel

I guess because something else loaded it or it's already com...

Yeah

the linux_load ones can be removed because the ABI service script knows how to load the modules

vmm_load probably should go in kld_list because vmm is pretty huge

Does it automatically load linux64 as well? I remember that one was being skipped for some reason.

on amd64 it loads both

On another machine I had to trouble with tmpfs that was really strange. It was giving me that already loaded error at boot, but whenever I started my poudriere jail, it would complain that tmpfs wasn't present. So starting it again after kldload tmpfs worked. It was weird.

...and yes, it was enabled in the jail config too.

the only way you'd need to load tmpfs is if you removed "options TMPFS" from the kernel conf

nvidia driver is another one to put in kld_conf for size reasons

snd_uaudio doesn't need loading since it'll be auto-loaded when a device is detected

That's what I would have thought, but I never touched it. The only custom kernel configs I used were when I needed to either enable ALTQ_* or add `nodevice ukbd` for my keyboard to work properly.

snd_hda can go in kld_list

actually it might not need to, devmatch probably can find it

Ah that's good to know. Slimming this down would help make it easier for sure.

if_*_load shouldn't be needed since the netif scripts know to load things like bridge and tap modules

getfacl /etc/rc.conf /etc/sysctl.conf /boot/loader.conf

if_re is the module from ports, needed it to fix a netif issue I was having with re0.

Title: FreshPorts -- net/realtek-re-kmod: Kernel driver for Realtek PCIe Ethernet Controllers

right, but if_tap, if_bridge etc. don't need explicit loads

Is there a way to specify the ports module from kld_list (if_re_load="YES" + if_re_name="/boot/modules/if_re.ko")?

kld_list can have the full path to the module

I think I explicitly loaded the other netif-related modules because I was using ng_*, bridge*, & tap* interfaces with jails. It's been a while, but if I'm remembering correctly, doing stuff like ifconfig ... vnet or ngctl ... from exec.prestart or whatnot was failing due to kernel modules not being loaded.

oh nice

You would think, based on what you said, that they would auto-loaded. Maybe I was doing something wrong, not sure.

I don't know about the netgraph modules.

I'm going to try it without loading those, see if it fails, and if so, what error it throws.

Is there an easy way to tell what is currently using a kernel module? (causing kldunload to fail)

Hmm, if kld_list is indeterminate as to when it's actually going to be processed during boot, then could putting if_re in there potentially cause netif to fail to initialize the NIC(s) if the kmods haven't yet been loaded by that point? (or would they definitely be loaded early? before FILESYSTEMS?)

s/initialize/configure/

yeah, if_re probably needs to stay in loader.conf

meena; Oh btw, not sure if this should be added to the pkgbase readme, but kldxref should be run after installing kernel package(s).

kld_list is loaded after kldxref and therefore after FILESYSTEMS, but it can be arbitrarily late after that

the whole business of ordering of rc.d scripts is a mess

Yeah, that part is somewhat annoying. Whenever I'm whipping up a port and doing an rc script for it, I have to grep through /etc/rc.d & /etc/defaults to get an idea of the appropriate header comments.

it's understandable to keep them weakly specified to allow for possible parallelism, but there are a lot of possible hidden dependencies that get violated

Lets just throw the whole thing in the trash and port systemd over to solve all of these issues from the 90s.

heh

I'd still rather deal with rc than systemd. Anytime I've had to do that in the past, it was a major pain in the neck trying to figure out the specifics for a given use-case. Also, the documentation felt a bit...under-emphasized lets call it...compared to BSD-land.

I'd settle for some way to add extra dependencies in rc.conf

man rc.conf

oops

Extra dependencies? Like service dependencies?

to be able to say "start X before Y" for arbitrary services

Ah yes, that would be really nice. Heck, that would even save me some hassle right now.

Something like nginx_service_depends="name0 name1 ..." would be nice, though I guess it should probably be limited to only being able to say "Services A, B, and C are required to be running before service Z" and not really support "Service Z should run after A, B, and C." (so then it can start with the root services and launch dependencies instead

of generating a list of services to explicitly start first and then starting whatever is left without explicit dependencies specified).

There's also the issue of running vs. healthy service :/ As in at what point are any of services A, B, & C considered to actually satisfy the Z dependency. Once it's launched and the PID is present for more than X seconds or leave it up to the requiring service to determine that itself.

Title: Start all of your commands with a comma

That's actually...a cute hack.

yuripv: thanks for the tip. I've tried a few extensions but twitch keeps complaining about recommended browser. I'll try to keep messing with it though

fancycade, My bank chase.com will not let me log in with a User-Agent defaulting to FreeBSD. I must override it and say that I am Linux and then it works.

yeesh

I use and recommend the Firefox extension "User-Agent Switcher". It works excellently. It can be told to automatically set the User-Agent on certain URLs like chase.com.

What does chase.com actually do when it's FreeBSD?

Freak out.

I wish sites would let user-agent sniffing die and just stick to feature-flags/capability-testing.

rwp: Could post a link for the extension. There are multiple "User-Agent Switcher" extensions

It will probably die in the near future. Chrome has scrapped the API that made it possible.

Which API?

Unless you mean User-Agent isn't going to be sent as a HTTP request header, which I can see causing a heck of a lot of issues.

Manifest V3 has removed the chrome.webRequest API. It would let an extension intercept requests, and insert the User-Agent header.

Oh, you meant via extensions like User-Agent switcher.

Yep.

Yeah, v3 is...*sigh* Lots of stuff not going to work anymore.

Wait, does v3 break any part of native host messaging? I made and use a couple of ports on FreeBSD that utilize it.

Mainly for org-mode capture and KDE Connect extension, both of which use native host messaging.

fancycade, addons.mozilla.org/en-US/firefox/addon/uaswitcher (and the upstream source is gitlab.com/ntninja/user-agent-switcher)

Title: User-Agent Switcher – Get this Extension for 🦊 Firefox (en-US)

_xor: I don't think so.

ox1eef_, Google is forcing the move to Manifest v3 but this is Firefox.

Firefox capitulated on v3, didn't they? I use Firefox 98% of the time.

Firefox is a clone of Chrome's API.

AFAIK Firefox is not moving to Manifest v3. But I am not an authoritative information source on this.

I will predict that when Google forces the move to v3 that it will create a lot of refugees to Firegox though. I bet they get a lot of users back again.

Title: Manifest V3 migration guide | Firefox Extension Workshop

Doubtful: extensionworkshop.com/documentation…develop/manifest-v3-migration-guide

Snap xD

heh

rwp: Right, it would have been wise to not drop support. It could attract more users.

I think I remember Mozilla originally stating that they weren't going to introduce most of the breaking changes that come with v3, but then I think I remember them reversing that later.

rwp: Thanks. That is the one I have installed. I can confirm on other sites that it changes my user-agent string, but twitch is still blocking it

That page is more than I can grok at this late time of night. If Firefox does go with Manifest v3 then it is going to suck for most of the users.

It made me trust Google even less. Why invest in developing for a platform where its APIs can suddenly disappear, and leave your business without any options.

I don't know about twitch but that leads me to think they are looking at something different from User-Agent then.

godaddy is the same...

I'm pretty sure they are. I got some announcements from some extensions that I used about it not working after so-and-so date due to v3 being enforced after that.

I bet the actual user market for extensions is pretty miniscule compared to the total user market. They're probably willing to make the trade-off.

It makes sense for Google being an advertisement business to block extensions from being able to block advertisements on security vulnerability basis. But Firefox has no such motivation.

I despised not being able to self-host browser push notifications either, but looks like VAPID allows that now.

_xor: can you not? i thought with ssl and a manifest that was possible

Last I looked at it a while back, you could control the messages themselves, but the actual delivery had to be done through service providers (e.g. FCM, APN, Firefox, etc).

Though now I think you can with VAPID keys.

_xor: I had a few ideas that I think could be monetized via browser extensions. And the native extension API opens a lot of doors. But who knows if it will live on or not.

I wanted to be able to fully self-host push notifications from server to browser, ios, & android. The issue was that browser push, at the time at least, required using a provider that was whitelisted in the browser by default. iOS & Android push was an issue because they required delivery through their services as well with the stated reason being

"power management" (so it wasn't reliable to just connect a WebSocket in your app and expect to always be able to receive, it would get put to sleep; only native Google background services would wake up and poll for you)

I wish the DOM-to-WASM bridge would hurry up and come to fruition. That would probably help solve a bunch of these issues.

oh, for ios and android. i know nothing about that, as i thought it needed a real app

Yeah. Browser restrictions can bite sometimes.

What do you mean by real app?

Oh a native app? That's what I meant, but you can still technically push via webapp on mobile using a PWA with Manifest + ServiceWorker.

Being able to cross-compile on FreeBSD to target iOS/arm64 and Android/JVM would be nice, but alas, my attempts at iOS failed hard and early (and bhyve doesn't support Clover or OpenCore to be able to boot macOS last I checked). I did get Android Studio working natively, but that was a major PITA.

RhodiumToad: I just checked the buildworld compile and it looks like it's actually compiling lib32 now.

I guess there's some dependency on DEBUG_FILES and/or KERNEL_SYMBOLS, because after including those in the build it looks like it's working (so far).

fancycade, You mentioned godaddy.com was the same problem. I just logged into a godaddy.com account using my FreeBSD Firefox with the default User-Agent and had no problems. Worked for me. Maybe something is simply broken with your Firefox?

Hi folks! I have a new Alder Lake-N system on the way

(see en.wikipedia.org/wiki/Alder_Lake#Alder_Lake-N )

Title: Alder Lake - Wikipedia

there's been a variety of patches to linux over the past ~year for this CPU family, since intel seems fairly proactive there--first being here, i believe: git.kernel.org/pub/scm/linux/kernel…9340326df40ba4462d4d2a1692b6387a68e

Title: kernel/git/torvalds/linux.git - Linux kernel source tree

but, afaict, this CPU family (0xBE) doesn't even exist in main yet for FreeBSD: cgit.freebsd.org/src/tree/sys/dev/hwpmc/hwpmc_intel.c

Title: hwpmc_intel.c « hwpmc « dev « sys - src - FreeBSD source tree

rwp: Perhaps. I installed firefox via pkg

(see 0x97, 0x9A at line 215)

I've tried all the work arounds. User agent, turned off tracking protection, private window, etc. Nothing works

fancycade, I also installed firefox via pkg. Currently at firefox-112.0_1,2 running on 13.1-RELEASE at the moment.

question 1: any idea what happens if i try to run freebsd on this, when the CPU type isn't explicitly defined in ther kernel? world of pain / poor performance/power, or mostly work? 2. what's the process for getting this added? (i started emailing the guy who authored the alder lake lines in that file, but that's probably not ideal)

rwp: firefox -v "Mozilla Firefox 112.0". 13.2 release, but I remember having similar issue on 13.1

I would suggest creating a new profile, to get something pristine, and then trying it. firefox -P

Hmm, didnt seem to help

I have all extensions removed at this point as well

It's a tough one. :-(

yeah, i've experienced something similar with alpine linux on this computer specifically too

thanks for the help though!

RhodiumToad: Oops, nevermind. Spoke too soon, lib32 failed at exactly the same spot again. Will tackle it later.

p14nd4: i would expect it to mostly work

Any clue why virtualbox stopped working after 13.1 to 12.2 upgrade? pastebin.mozilla.org/8RNamVv8

Title: Mozilla Community Pastebin/8RNamVv8 (Plain Text)

Gpart's error messages could be improved: paste.debian.net/plainh/6870812d

Got MariaDB 10.6.12 installed under FreeBSD, I see the following Warning line in log file; 2023-04-20 2:35:00 0 [Warning] You need to use --log-bin to make --binlog-format work.

any clue?

updated pastebin.mozilla.org/4BsTybXP

Title: Mozilla Community Pastebin/4BsTybXP (Plain Text)

`file` reports file system labels with appended whitespace: paste.debian.net/plainh/b5e9267d

OEM-IDs are also affected.

Looks like undesired behavior to me.

is that a bug or a feature?

That's the question.

I can't see how this would be a feature, but maybe other people can.

Hello, my laptop (Lenovo TP p14s Gen1 with FreeBSD 13.2) rebooted by itself without a warning and now 2 of my datasets are just gone. I mean, dataset are still present but there are empty, no data in it !

How it is possible ? Is it recoverable ? How can I investigate ? I read the zfs man pages and various thread on forum but found nothing relevant to my case

more over, how prevent this to happen again (I may have badly created these datasets)

datasets are empty or mountpoints are empty?

those datasets where created by "zfs create zroot/mount/point/ddataset", no extra options

(are they mounted?)

yuripv: I can't tell as I'm new to FreeBSD. I suppose mount points are empty but I'm unable to retrieve any trace of the datasets

do they show up in zfs list?

yes the show in zfs list and in "mount"

what does the used column say?

so I suppose data are gone but I can't tell how data can just disappear

is the mountpoint where you expect it to be?

futune: the used column says 96k for each of them :(

yes the mount point is where I expect to be

doug713705[m], oh... if there is no data used then yeah

my guess would be that they were not mounted when you placed the data in them

I have copies of the data so it's not a big deal by itself but I'd like to understand how the data have vanished and how to prevent this behaviour !

unmount the datasets and check what is in the mountpoint under it?

ok, so umount should make the data reappear

ok, lets try this, sounds like a good idea

umounted the datasets and no data :(

ok, I don't have any further ideas right now and I have never seen this behaviour before...

Ok, so I may have hallucinated and never put the data in the datasets

did the machine reboot within 5 seconds of writing the data?

zfs usually writes to disks in 5 sec intervals

no, the reboot occurred several days after data copy

so, again, I may have hallucinated... I will check my logs and verify that I have really rsynced the data in the first place

hope you figure it out!

there is also #zfs if it really appears to be a zfs problem

maybe someone there knows more

Title: BSD Now 503: Fast Unix Commands

futune: II have no trace of rsyncing the data /o\ A good old pebcak

doug713705[m], that is reassuring to me... maybe not so much to you :(

yes it reassuring, at least for my system health, not so much for my brain /o\ :D

:D

This didn't helped either forums.freebsd.org/threads/freebsd-13-2-and-virtualbox-ose.88747

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

Is 13.2 broken ?

Beladona: in what context?

virtualbox, kernal issue

for one example

Title: Mozilla Community Pastebin/4BsTybXP (Plain Text)

so instead of asking why you can't build the kmod (even if the error message tells you why) you are asking if 13.2 is broken

i got a great idea: We announce the availability of 13.2 3 months after it's released, and thus, 13.1 has been deprecated.

Asking questions is hard. I've been doing it for 40 years and I still often get it wrong.

yuripv what can I do?

What does the message say? what is the problem with virtualbox?

If you ask me, virtualbox works just fine...

but what are you trying todo specificly?

Kernel driver not installed (rc=-1908) Make sure the kernel module has been loaded successfully..where: suplibOsInit what: 3 VERR_VM_DRIVER_NOT_INSTALLED (-1908) - The support driver is not installed. On linux, open returned ENOENT

KLD vboxdrv.ko: depends on kernel - not available or version mismatch

linker_load_file: /boot/modules/vboxdrv.ko - unsupported file type

drobban this

Title: Mozilla Community Pastebin/4BsTybXP (Plain Text)

Beladona: "/usr/share/mk/bsd.sysdir.mk" line 15: Unable to locate the kernel source tree. Set SYSDIR to override.

ya, how exactly?

something sound like you missed something in your upgrade

am.. not sure

 I even did `pkg upgrade -f`

Havent upgraded to 13.2, My world started with 13.2, and virtualbox works just fine on my machine(tm)

Beladona: do you have /usr/src with the 13.2 source code?

checking

so perhaps the driver is compiled with older kernel in mind...

and you need to update kernel source on your machine...

# ls /usr/src/ shows nothing

so the message made sense then

ok, how can I do that?

perhaps downloading correct version will solve your problem?

Beladona: how did you upgrade?

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

is this relevant for your situation?

I just did freebsd-update upgrade -r 13.2-RELEASE

I will follow this now docs.freebsd.org/en/books/handbook/…updating-src-obtaining-src-repopath

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

that is probably what I would have done... But hey, I have never updated system in freebsd before -- Im a freebsd n000bie

no problem. Doing. will let you know

super

yuripv: do you want lspci -v or -vv ?

heads up ... src/bin/pkill/tests/pgrep-F_test.sh conflicts with src/bin/pkill/tests/pgrep-f_test.sh on case insensitive systems causing the repo to having problems with updating... can anyone get some eyes on this? ... its a real pita while working on case insensitive filesystems...

CmdLnKid: ugh, which system you on?

this is weird, cuz surely we have people working on macOS; like, the whole of buildworld and buildkernel compiles there

i work with the repository on macos "apfs" which is default case insensitive

easy enough for me to work around but really honestly shouldn't have to be that way

is there any fun benchmark software out there where I could compare results with other peoples computer setups?

CmdLnKid: can you open a bug, and add a patch to rename those files

dunno about a patch as the generated patch wouldn't reccomend anything other than file1 -> file2 renames.

i can file a bug report though

pgrep-F… → pgrep-Pidfile_test.sh ; and pgrep-f… → pgrep-full_test.sh

Title: 270948 – Conflict with bin/pkill/tests/pgrep-F_test.sh / bin/pkill/tests/pgrep-f_test.sh

what kind of benchmark?

paulf: whatever... hardware in general

im just trying to waste time here.

may be better to state your goal

get some random comparable results?

but your end result is ...

is it possible to incrementally compile the kernel?, I just modified a driver that's builtin into the kernel but need to rebuild the whole thing again, and my systems takes like 1 hour to compile

neirac: can't just cd into the driver and do make?

neirac: Meta mode, maybe?

is this a module? or a builtin one?

meena builtin one

vkarlsen what's Meta mode ?

neirac: wiki.freebsd.org/MetaMode -- "[...] an incremental build system for FreeBSD, which uses bmake's meta mode and filemon(4) to create metadata files, substantially decreasing build times by only building what has changed since the last build"

Title: MetaMode - FreeBSD Wiki

vkarlsen thanks!

neirac: I hope it helps, it's been a while since I used it myself

vkarlsen it only works with buildworld or also with buildkernel? according to the doc only refers to buildworld

neirac: I don't know

by the doc it seems pretty clear its buildworld but by the author that may have been just a quote as being everything

just an assumption

pretty cool none-theless

ill give it a shot on a RO nullfs mounted from a zfs

*.meta files are also present in /usr/obj for kernel, so i would assume -- yes, it should work for kernel too. meta mode that is

keeps my source tree clean as i learned in the past for some reason object files were left behind while compiling on a RW filesystem even tho a obj filesystem was supposed to be used.

now i don't have to second guess my source tree

fstab: /export/usr/src /usr/src nullfs ro,noatime,nfsv4acls 0 0

they are files automatically created with meta mode. i do not know how unlcean it is as it upposed to be that way

ill find out in a moment

;-)

appears fine so far ... find /usr/obj/ -type f -iname \*.meta

find /usr/obj/ -type f -iname \*.meta |wc -l

175

next build after this im assuming it will be twice as fast

or at least faster

completely unaware there was src-env.conf. been a while since i revisited that

last i know it was src.conf only

s/know/known/

what would be the best approach to export several disks as a single LUN (iSCSI) other than a zvol?

Hardware RAID an option?

iSCSI is block-level storage, so it'd have to be seen by the OS and one contiguous device, prior to export via iSCSI.

nop, I'd like to use ZFS on the client-side

and I can't find the information wheter $> zpool create mirror zvol1 zvol2 is supported

test it out on some local file backed zfs storage

dd if=/dev/random of=some/path/file[1-N] count=1000 bs=1024

zpool create ... after you attach those to a md device

reverse the process to delete them

meena: lspci -nn please

I think you're confusing OSes. . .

pciconf -lv

common man -k conf answer

;-)

CrtxReavr: were comparing outputs from fbsd and Linux

Not allowed.

mage: "nested" zfs didn't really work for me ever

CrtxReavr: I'm not looking at GPL code, just at its output ;)

yuripv: updated

thank you

meena: so yes, it's the same in linux (meaning not a FreeBSD bug), only there it says "unclassified" instead of "old"

hrm…

mage: layering zfs on top of zvols isn't going to do anything good for you, even if you know exactly what you're doing (and if you did, you wouldn't be doing it :P)

ctld(8) using a zvol as a device extent is the optimal solution

Apparently, `umount` is unable to collapse consecutive slashes in device names: paste.debian.net/plainh/594a6076

fix it!

no c skills!

I've worked around it for now by making use of the venerable tr(1).

umount expects to be told the mount point, not the device name

if you just give it a device, it has to try and find it in the list of mounted filesystems, which is a text match rather than a matter of resolving inodes

Well, but the man page says, gving the device name is just fine.

Also, how would a text match prevent it from collapsing slashes?

how do I get the sound to work

on what hardware?

vmware works with root

what do you mean?

sound doesn't work with kiwichap

but does for root?

yes

check the permissions on the /dev/dsp* devices

how?

first, what happens if you run the command 'beep' ?

nothing

it completes successfully? or you don't have that command?

on snd_hda how do I get sound through headphones?

brb phone

that depends on how the hardware is wired.

in some cases, plugging the headphones in is enough. In others, the headphone output is on a different pcm device

sometimes you can tweak the hdaa config to handle it.

RhodiumToad thanks, I'll go read snd_hda then seems needs some tweaking

usually you can find the headphone output either from /dev/sndstat (for some easy cases) or from looking at the hdaa geometry

for example, sysctl dev.hdaa.0 will show you the pins and connections for hdaa device 0

if I try to kldload SGX and gat this error: `module_register_init: MOD_LOAD (sgx, 0xffffffff82c25000, 0) error 6`, does that mean the CPU doesn't support SGX?

typically, yes

if SGX isn't supported that's all you'll get. there are some other cases where you'd see another message as well

you should see SGX in /var/run/dmesg.boot if it is supported:

Structured Extended Features=0x29c6fbf<FSGSBASE,TSCADJ,SGX,BMI1,HLE,AVX2,SMEP,BMI2,ERMS,INVPCID,RTM,NFPUSG,MPX,RDSEED,ADX,SMAP,CLFLUSHOPT,PROCTRACE>

hi sorry back for 15 minutes

sorry it was the welfare office

how do I get the sound to work?

on my user account

beep never ever works on anything

I did mixer it's connected it's on vmware , it works with root

sorry call again with welfare

thanks RhodiumToad, ye, its not in the list

hi thanks RhodiumToad

sorry about that I am trying to get the sound to work

how do I do that?

does 'beep' work as root?

never works on any device

and what exactly _does_ work?

root : youtube, or like any sound works generally but not beep

user: kiwichap : no sound whatsoever , other installations such as laptop bare metal installation, sound works with no issues

so in a browser it works?

yes as root

in which browser?

and if it's firefox, what does about:support say about the sound config?

ungoogled chromium browser but I believe it's the same as otter

there's no beeping on the terminal etc..

does chromium have an about:support or anything similar?

it's the user account that doesn't work I tried on otter browser it's the same issue

I can't help you if you can't answer questions

yes chromium has settings

what I need to know is: in the configuration that works, i.e. running a browser as root, which audio backend is the browser using

support, not settings

and there's also a group nntp on chromium , they have forums, there's youtube, mailing list

I don't use chromium. in firefox, if you go to about:settings and scroll down to "Media", it shows what audio configuration is actually in use.

what I want is either that info, IF IT WORKS, or the equivalent for chromium

interesting, it doesn't have this

if you want I can log out and login as root then test sond

sound

I don't want anything. If _you_ want help, you must answer the question.

what question there is no info

no there is nothing like that

in which browser?

in otter browser and ungoogle chromium

I. Don't. Care.

ok

so should I install firefox I'll do that

you need to find out how to tell which audio backend is in use in a configuration where audio works. Don't bother me again until you have that info

I have done that with cat /dev/sndstat

it says no devices installed from userspace

nothing else?

then you have no sound hardware available (real or emulated), so the only way you could have been playing sound is via network, e.g. with pulseaudio

which is why I was asking about audio backends

so you have a pcm0 device.

what exactly happens if you do beep -d /dev/dsp0

would have started to test cat /dev/random >/dev/dsp

no sound

with your user and with root?

I have to logout but with root I get sound

you get sound when using one of those commands?

not sure I didn't use them but if I go to youtube with a browser I get sound for example

or if I click backspace on terminal

so it seems like permission problems if that cat work from root but not from your user

backspace on terminal is a different kind of sound, ignore that completely

can you make sounds using /dev/dsp when logged in as root, is the question

I will logo out and login

be right back

hello I am back

doas!

substitute user

just login as root

on your serial console

right, I wanted to look into that new fangled doas stuff at some point

it's already old and maintenance free

I use doas exclusively for quite a while. I only have sudo setup for system administration utilities that need advanced capabilities not offered by doas (such as sysutils/cbsd)

the benefit of sudo is better support in third party libraries and application extensions (modules) due to its age and adoption

exactly. for anything else -> doas :D

what is the advantage of using doas?

and a stronger developer skill behind sudo

doas is much simpler, lower complexity, easier to configure.

Let's not have that discussion again.

todd miller vs ted unangst

its eems I need to have sudo set up still, so "less complex / easier to configure" doesn't really apply

but ok

Stop. Just stop.

Let the Linux zealots fight the conversion wars.

shall we talk about the one true editor now? :p

Take a guess.

there is no built-in editor for doas.conf ;-)

Well, visudo isn't an editor, it's a configuration sanity editor - which isn't a bad idea, when a misconfiguration can lock you out of your system.

err, configuration sanity checker*

It even respects the various environment variables related to editors.

ye, I used it with mg

if only mg was 8bit clean

there is probably a way to just use emacs tramp mode with visudo

tramp supports doas just fine

no comment

is there a good reason the rpi images don't ship with zfs root?

for the newer arm64 pis

yes

care to elaborate? I don't think performance would be any problem

it handles btrfs with compression on luks no problem :D

that's not important, the important part is, emacs is not unixy

and neither is python, so there is your "pi" correlation

Right. It's not unixy. Nothing other than ed is unixy. Since ed is the STANDARD editor. Says so right in the man page!

that was about sed

In any case visudo may edit files with emacs okay. No problem! :-)

I looked at doas and saw some reason not to use it, but I forget what it was

sudo alternative from OpenBSD

sudo also lives in OpenBSD

I started using doas recently just to check it out. It's less powerful than sudo, but also simpler.

The syntax of /usr/local/etc/doas.conf is easier to remember than /usr/local/etc/sudoers.d/${USERNAME}

I can understand people advocating for doas because they like it better. I do not understand the opposition to sudo.

conf.d/* style is becoming more popular for good reason

there are use cases for both, based on functionality, not artefacts

_xor, What is /usr/local/etc/sudoers.d/${USERNAME} supposed to signify? That would just be another file included. Nothing special about ${USERNAME} there.

I generally don't expect to get actual answers (though I'm willing to listen) from zealotry.

rwp: What? I was just referencing the syntax of the files in sudoers.d/.

On systems where etc/sudoers is not maintained by a package manager then I use it and configure it directly. On those systems I use etc/sudoers.d/local-sudoers for my local configuration instead.

epony: yes, but everything that isn't OpenBSD gets the portable version of doas.

hurray for PAM-less systems

RhodiumToad: I found it somewhat odd that the rc framework supports *_env and *_envfile but not much beyond that. I have a template rc script I wrote a while back that uses ${PREFIX}${ETCDIR}/profiles.d/ to load environment variables if running multiple instances of a service (e.g. haproxy).

Well, maybe not odd, but I guess there wasn't enough of a need for it. Instead of needing to do something bespoke like how net/haproxy does ${name}_profiles="site1 site2 ..." and the rc script implements handling multiple profiles that way, it makes more sense to me to have multiple profiles specified as environment variables.

When I see things like profile.d/foo then I think, the user must be asked to log out and log back in again for the profile to be loaded into their environment. Would prefer to avoid that. I know I can manually source the profile and on down the environment manually.

Though I'd be surprised if multiple instances of an rc service is an uncommon requirement.

Instead I very much prefer the style where a script loads the current environment configuration if one is needed on the fly. That way it can be changed and upon the next invocation the script gets the updated environment immediately. It's simpler to invoke that way.

_xor, And example for your point is OpenVPN might be used for multiple independent connections.

in that regard doas lets you control your environment

As just one example I have been in the middle of recently. On a, ahem, blech, a systemd system.

Where would I submit a patch for review, and possible merging, if I copied my rc script over to /etc/rc.subr and possibly /etc/rc.d/* to use the optional approach I describe above? Phabricator?

Wouldn't be too difficult and might be useful to others, though it's a bit of a mess so would probably need to be cleaned up. I can see it being done by adding 1-2 more functions to /etc/rc.subr and then possibly updating /etc/rc.d/* scripts to utilize them (along with relevant rc knobs) where multiple instances of the given service are run.

there's a few services where multiple instances make sense, and they currently tend to do their own profile hacking in the rc script

dhclient?

ppp is the one that comes to mind

Does dhclient just check for ifconfig_* knobs and then filter out the ones with DHCP in the value and process the ones that do?

dhclient is per-interface, iirc

so it's being specifically invoked with an interface name, not just generically started

Right, but is there a single instance of dhclient that gets invoked to handle all DHCP-enabled interfaces or is it one-instance-per-netif? I want to say I've seen multiple dhclient processes via ps, and so I'm guessing the latter.

I imagine the rc script for dhclient filters a list of ifconfig_* knobs, and then loops over them to look for DHCP somewhere in the value and launches dhclient for each one it finds.

no

the rc script for dhclient literally has an interface name as a parameter

oh, well then <_< >_>

Heh, I tried to grep for DHCP and didn't back much. Then I ran it on /etc and saw that network.subr has a functions to check for DHCP and WPA.

Was mainly curious to see what/where/how it is that handles ifconfig_${NETIF}="... DHCP ..."

well that's not at all obvious

network.subr runs /etc/rc.d/dhclient start $1 for interfaces configured with SYNCDHCP

Agreed, not obvious at all.

the dhclient script has keyword nostart, so it's not executed itself

That seems a bit...disjointed.

I mean, it's not executed directly as part of rcorder

Probably missing a legit reason as to why, but it makes more sense to me to have something like ifconfig_${NETIF}_dhcp="..."

When building a port, is it possible to have pkg install any binary dependencies for that specific port so that the ports system doesn't also build all the dependencies?

My router has an interface that needs to use DHCP and also set a static IP. It wasn't obvious to me at first for some reason, but I figured out ifconfig_em0="SYNCDHCP" + ifconfig_em0_alias0="inet ..." did the trick.

Though I think it would be nice to be able to do something like ifconfig_em0_dhcp="..." to set explicit flags and/or config file on a per-interface basis. Or heck, could even do it by convention where it looks for the presence of /usr/local/etc/dhclient.d/em0.conf or something and uses that as the config file for dhclient.

Never mind.

iio7: Synth does that, don't think poudriere supports that IIRC.

latest poudriere does, I believe

haven't used it yet

Really? :O

if you're just building the port with make, there's some support for using pkg add (not pkg install) for dependencies

I assume you're talking about poudriere-devel?

yeah

Looking now. If so, that should save me a good amount of time.

Oooh, lookie at that pretty -b <branch> argument.

Does anyone know what has happened to the signal-desktop package? It is suddenly no longer present when searching via pkg, but it is in net-im in ports.

Hmm, interesting to note about custom knobs.

Any idea what commit this is? bugs.freebsd.org/bugzilla/show_bug.cgi?id=221122#c28

Title: 221122 – Attaching interface to a bridge stops all traffic on uplink NIC for few seconds

After an upgrade, the package was suddenly gone (I didn't notice during the upgrade, until I needed to use the package).

iio7: You sure it wasn't removed during pkg upgrade? Are you using both packages and ports together?

_xor, no, only packages.

_xor, I believe it was removed, without my noticing. I used it just the day before I did the upgrade.

Is there a pkg log where one can see what has been added/removed?

You can look on both freshports.org as well as portsfallout.com

Title: FreshPorts -- The Place For Ports - Most recent commits

pkg logs these things to /var/log/messages

...and that

Nice. And yes, it was removed, for some reason. "pkg[60565]: signal-desktop-6.12.0 deinstalled"

But also, doing a search with pkg doesn't show the package any longer.

iio7: I'm looking right now, one sec. Port appears to fail in configure.

Oh wait, no, it fails in build not configure.

iio7: I can't finish looking at this right now due to time constraints (the port has to fetch a bunch of files and it also uses electron, which is its own can of worms), but it looks like the port was updated within the past week and may address the issue you're seeing (specifically, it's not copying a patched file required for a dependency due to

an invalid destination directory).

iio7: If you want to wait, it may build in the next run, or you can try building it yourself after updating your ports tree.

_xor, I believe there maybe is a problem with electron22 which maybe the reason why it was removed.

_xor, thanks.

Could be, wouldn't be surprised. Electron is annoying.

_xor, yes!