uff, we still have stuff depending on llvm10?

so that means id have to install llvm12 to 16

unless i really need newer llvms?

firefox-112.0.1,2 llvm13-13.0.1_3

jbo: oh you mean the feature i've been waiting for on gentoo for years? nice :D

well, ports depend on stuff (both for building and actually running the resulting binaries). some ports require different llvm versions. hence you want to fetch them rather than building them yourself.

ixmpp, welcome to the world of freebsd ;-)

although i wager my adding debug flags would preclude that

i c

ixmpp, exactly. BUT poudriere will be smart enough to figure that out.

another feature that portmaster lacks :p

hmm

jbo: not sure about llvm10. I thought lang/ghc was going to need it, but it's not in the deps list.

so portmaster is outdated, poudriere all the way?

RhodiumToad, if you have the poudriere dashboard available it should tell you why it is building llvm10

for anybody new: poudriere all the way.

I know why, it's because I put it in the list :-)

RhodiumToad, I didn't dare to say/ask :p

welp, found out my pc was using 12 gigs out of its ram and it was all firefox's fault

RhodiumToad: Nice, LLVM is like Pokemon - catch them all!

was it firefox's fault or the fact that browsing a modern web page requires 1GB of memory?

mason, :D

it then crashed and ram usage wnet down to just 4gb

and yeah i was with like a dozen tabs of Modern Websites

doesn't sound like firefox's fault then.

ISTR at one point FF (for a single process) showing memory use larger than host RAM+swap.

it occurs to me that gentoo's slots system is a steep overcomplication when i think about how ports does things

it looks nicer though :p

ah. ghc requires llvm10 on arm

but apparently not on amd64

ixmpp, so you're a gentoo user? are you ready for a fun read (but slightly painful)?

ha, sure

Title: Welcome to Gentoo is Rice, the Volume goes to 11 here.

oh god

jbo: Was that his originally or did he just copy funroll.loops?

(or whatever the site was)

ok i'll take the hit, but to be clear, i don't think anyone uses gentoo for speed

:D

mason, no idea. I haven't been around for long enough. afaik it's just quotes the person collected over the years from fora, irc etc.

jbo: that used to be a standalone web site

ixmpp, I don't judge - just wanted to share what gave me a good laugh :)

mason, that sounds like fun

jbo: web.archive.org/web/20080702015125/http://funroll-loops.info

Title: Welcome to Gentoo is Rice, the Volume goes to 11 here.

Worth archiving, of course.

mason, really no idea - might be the same dude with different website. I really don't have a clue.

@mason archive everything you find

I'm going to go archive dinner, right now. o/

well, i dont have space to archive it all so i just archived anything i find involving parappa the rapper, but archiving a few things is better than not archiving

and cya

it does have some gold in there tho

> Yea, I really don't understand all the complaints about the time to install gentoo. It is like complaining about your Ferrari because the dealership was so far away.

also, forgot, does poudriere have a feature to apply patches to packages? thats a feature of gentoo i absolutely love

that and savedconfig

if you want to do your own patches, you put them in the port

but both can be done better

it's easy with git to maintain your own branch of the ports tree with your own changes

RhodiumToad: oh! and because theyre untracked - yeah

i knew you guys would do it better somehow

ixmpp, patches are handled by the port system itself.

aye, too slow. sorry.

this is why i keep wishing i could use this on desktop, but there is so much friction with the hardware...

ixmpp, not everything we do is better tho. we just try really hard whenever possible :>

but freebsd is the way

maybe you can get a cheap 2nd hand replacement laptop? Lenovo Thinkpads are generally very well supported.

might be realistic eventually

I'm trying to run a Linux binary but it's complaining about not finding ld-linux-x86-64.so.2 : do I need to setup the CentOS Base System as well ?

right now I just did service linux start

yep, that was it: pkg install linux_base-c7

jup

the linux service only starts the linuxulator layer

if your linux binary depends on other stuff you have to take care of that

my app does come with some library dependencies, some .so files that shipped with it

I placed them in /usr/local/lib/solidigm and ran ldconfig -m /usr/local/lib/solidigm but I guess that won't work for Linux libs ?

for linux libs you have to run the linux version of ldconfig

chroot into the c7

then do ldconfig in there

you don't need the chroot, but you might need a -C filename option to make sure you update the right file

the libraries should probably go in /compat/linux/usr/... rather than under /usr

I have a linux printer driver, what I did was just stick the .so it needs under /compat/linux/usr/lib64

yep, I did that, but it turns out I also had to guess the correct subdir

so in this case it wanted /compat/linux/usr/lib/solidigm

it might not be enough. this is the Intel/Solidigm SSD firmware update tool

on Linux I can get it to work by specifying -ssd X , aka -ssd 0 for the first drive

on FreeBSD it says: Error: Invalid drive identifier , maybe it can't run under emulation ?

oh, that will surely require some more magic.

might be better of just running a live linux system and upgrading firmware that way.

i wish lldb was more comfortable to use than gdb

?!

but for me at least, nope

ixmpp, lldb.llvm.org/use/map.html

I'm trying not to use Linux at all :)

Title: GDB to LLDB command map — The LLDB Debugger

last1, if you use the linuxulator you are using linux x)

linux-c7 is literally a linux base

but it's probably looking for different devices in /dev or something ?

wow, theres more than i thought there was in lldb

ixmpp, what did you expect? llvm is a fully fleged compiler suite. just like gcc. but different.

might give it a try again someday

i've tried it before

it was ...painful

seems like a problem on your end :)

well, granted, it is like learning a whole new language

it's like learning a whole new CLI.

ok yes

this just seems like typical linux mentallity to me (really, no offense at all)

"it's different therefore it's bad"

and that from a community that is as fragmented as linux :o

no, im not saying its bad (anymore, now i know its as feature complete)

but it will still be hard to relearn

imagine i took vim from you

and you had to learn ...kakoune(?) or some other nonsense

so? I'd just learn whatever you give me instead. I'd start by reading the docs. I would not just blindly assume that it can't do a certain thing and throw that oppinion out in the world.

would hurt, right?

no. it wouldn't. just read docs, adapt and then make your mind up.

I might find that kakoune does something nicer than vim

im fairly sure i did at least try, when it came to lldb, but like you said earlier today(!) if it's not obviously better, why switch? and the gdb cli is more visibly complete, so do you blame me?

well, I get your point, but you're misusing my statement in a different context. here the difference is that if you want to use freebsd you better get used to clang because that is the base compiler :)

which is why im reconsidering it :p

you're free to choose. we have plenty of linux folks coming by because they get fed up with linux for various reasons, consider, maybe even try freebsd and then move away. some stay, some go away. that's everybodys freedom :)

if you want to stay & learn, the community will gladly help. if you determine that it's not for you - that's fine too!

I certainly don't miss GCC the same way I don't miss systemd :p

i've been using freebsd some time

on server only though

that's fine too

but i do still use linux, and by extension gcc

which is fine too

anyway im here to stay, just been more active lately cause linux crap has driven me to want freebsd desktop more and more

that is fine too :p

speaking of fragmentation, there's that chimera-linux too... fbsd userland, llvm, linux kernel, specific opinions...

that sounds like a nightmare to maintain

I'd argue it's a symptom of the underlying problem :p

just use (Free)BSD and use linux emulation or a VM if needed?

both nice & funny to see that it is using pipewire tho

moonshine: there was a gentoo/kfreebsd, and i think the debian one still exists

arch too irc

yeah the debian one still exists.

iirc

my server survives fine with bhyve vms

but thats cause all it needs is network and a userland, really

while we're on about the linux->bsd migration, the one thing that seems the most frictionful (aside from the change of toolchain) is that the bsd coreutils are much stricter on argument order

i'm actually not complaining, cause that is just something to get used to

but its interesting to note

i wish i'd started on a bsd now, because it works in the other direction, but not this one

(actually, my first non-windows OS was netBSD, but only briefly)

to be precise, the BSD utils don't follow the GNU-ism of allowing arguments to precede options

so it's command [options] [--] [arguments]

yeah

the GNU style can be very annoying from a shell scripting point of view

(since you have to worry even more about substitutions starting with - than in the normal case)

anyone used zpool remove command ? (man.freebsd.org/cgi/man.cgi?query=z…ion=8&manpath=freebsd-release-ports) and can comment on how well it works ?

Title: zpool-remove(8)

RhodiumToad: I dear god, I've been having to deal with that quite a bit lately.

I despise it.

*Oh dear god

I actually use this pretty heavily these days because it's at least consistent and cross-platform, plus adds some functions useful for scripting...

Title: GitHub - go-task/task: A task runner / simpler Make alternative written in Go

It integrates this lib, which works across FreeBSD, Linux, macOS, Windows, etc...

Title: GitHub - mvdan/sh: A shell parser, formatter, and interpreter with bash support; includes shfmt

Oh yeah, btw, do poudriere overlay trees still work without using ZFS? (maybe it uses unionfs or something instead of doing a zfs overlay mount?)

what's a good gcc version these days?

I've been using poudriere overlay trees, but might have to switch to a single tree with multiple branches that get rebased instead of just using overlay if non-ZFS poudriere doesn't work with them.

Don't know what's considered good, but I've not had too much trouble using 12.x whenever something requires it.

I haven't tried overlays myself

I've been using this for quite a while and love it: github.com/jdxcode/rtx

I just have one ports tree

Title: GitHub - jdxcode/rtx: Runtime Executor (asdf rust clone)

There's a port for it now too (had an internal port I whipped up, might get rid of it in favor of upstream, though this project has a new release like every other day, so to keep up I might stick with my own port): freshports.org/sysutils/rtx

Title: FreshPorts -- sysutils/rtx: Runtime Executor (asdf rust clone)

Works really well though for generically installing & switching between side-by-side multiple-version installs of apps (which includes gcc in my case too).

User installs too, so root perms are a requirement.

Re: Overlays, I like them because I just had fewer merge conflicts to deal with when updating main tree ports to new versions or making minor fixes. Just had to keep an eye on updates as they became available.

Oops, root perms are NOT a requirement is what I meant.

There's probably a better way to do this or a port made to switch between versions of LLVM, but I also use rtx when switching between LLVM versions (if required), since it's installed in /usr/local/llvm${MAJOR_VER} by default so as to not conflict with the base install.

I had trouble with clangd, though I didn't really look for the C++ equivalent of javavmwrapper. It was just easier to do `rtx install llvm@${VER}` and then either create .tool-versions (or .rtx.toml) or do `rtx global llvm@${VER}`

I also use it for PostgreSQL to have 13.x, 14.x, & 15.x installed. Helps when doing major version upgrades and if pg_* are needed from both old and new versions.

sounds nix-y

but without the pain of having to use nix

I tried nix, but it was such a PITA.

(-os)

Yeah

rtx is a clone of asdf. I tried asdf a long while ago as a replacement for pyenv/nvm/etc, but it's implemented in bash and was pretty Linux-centric, so the trade-offs weren't worth it.

rtx is an asdf clone with additional functionality and implemented in rust (yeah yeah, I know)...but it has worked beautifuly for me so far.

another thing im glad for in freebsd hah, the rust fanaticism isnt present in this community

i dont hate rust, but the hype is unreal and its not imo remotely worth it

I like rust quite a bit, but that's for my usage of it. It's a trade-off like anything else, and so yeah, the evangalism can get annoying.

veg, Good luck with your array!

ntpd doesn't stay running gon 13.2-release

there's not log message in /var/log i can find about this

anybody else experience this?

sfox: what settings do you have in rc.conf for it?

sfox, Are you using -g with it? See the ntpd_sync_on_start rc.conf variable.

sfox: in particular, have you set ntpd_sync_on_start?

just the stuff from the installer, ntpd_enable="YES"

Add ntpd_sync_on_start="YES" to solve the problem. That's normally needed.

without ntpd_sync_on_start, ntpd will exit if the clock is wrong by more than some threshold

ok

can i fix this without rebooting the server?

with it, then as soon as ntpd is synced up to an external source, it'll step the clock (once only)

sure

Yes. Set the variable. Then restart ntp. "service ntp start"

oh

Sorry, it's "service ntpd start"

ntpd_sync_on_start="YES"

right?

Yes.

btw, it's good to make sure that you have "iburst" set in ntp.conf for your best timeserver entries

thankyou

i just left it at the default when i installed freebsd 13.1

it has iburst

that's good

+1 on iburst but that should be the default already. "pool 0.freebsd.pool.ntp.org iburst"

Also, the documentation for "man rc.conf" then search down for /ntp and will see all of the options along with ntpd_sync_on_start we are recommending.

thankyou

RhodiumToad, It was fun teaming up with you on ntpd -g! ( o/ )

cange of gears for a second, i wanted to ask what the general consensus was on iocage

with it now being unmaintained, there doesn't appear to be a replacement for it yet

jailer explicity says not to use it in production

I really need to update my jails to the latest minor release 13.2

iocage upgrade is broken

I know M. Lucas recommends iocage in his FreeBSD books but so far I have not used it myself yet. I like working down in the dirt.

I can't do it inside the jail as it complains about the security level

assuming I can lower the security level of the jail and use freebsd-update without further issues, would there be any problem doing this manually this way

and updating the config file afterwards by hand for the jails

rwp, I read Lucas's book too

"inside the jail

I ended up migrating from ezjail to iocage

"inside the jail" confuses me as iocage is used *outside* the jail to manage the datasets.

i mean iocage console jailname, and from a shell running inside the jail run freebsd-update

So it is the freebsd-update inside the jail which is failing? What are the errors you are encountering?

iocage upgrade -r 13.2-RELEASE mumble works, but it has me having to resolve merge conflicts of every update file in /etc/

files that should just be replaced

but there's no option to replace them

That does not seem like an iocage problem as such. Can you console shell into the jail and run freebsd-update from the shell level?

I haven't upgraded to the new 13.2R myself yet. Makes me want to try it immediately in a victim VM and see if I have similar problems.

rwp, no when i do so it says the security level is too high to run freebsd-update

Thank you for the paste. The devil is in the details and that shows them most clearly.

I admit that this is beyond my day to day level of knowledge. I have so far only used jails here and there and I don't know enough here.

The security level is tickling some brain cells but I think I have not restricted my own jails such and so have not run into this problem myself before.

i don't believe i did any additiona hardening of the jails

just set them up as 'thick' jails

without vnet

the good news is that iocage zfs snapshots your jail before upgrading

I know there is also the "freebsd-update -j jail" option too. I am the blind here leading the blind maybe but perhaps that would upgrade your jail for you? Warning! I don't know what I am doing here.

so i can roll back to a consistent state and not have a half-way-inbetween upgraded jail

i'll try that and report

iocage seems popular, but bhyve seems simpler for the same perks, am i wrong?

it's doing it's thing now: # freebsd-update -j 11 upgrade -r 13.2-RELEASE

ixmpp, isn't bhyve virtualization not jailing?

bhyve is full virtualization, yes

oh sorry, im thinking of ...what is it, iohyve

i've read online that FreeBSD administrators typiclly avoid upgrading to -p0 releases waiting for a -p1

maybe that is what I should have done

misinterpreted, ignore me

crap

it broke this way too

Well... It's 13.2R so already past the .1 release. So probably okay on the p0 front. But this does not "feel" like a p0 problem but more like something else.

rwp, paste.debian.net/plain/1277586

Hmm... I should create a victim VM and then walk through an upgrade to 13.2R. Sorry. I haven't done it yet.

it's currently at 13.1-RELEASE-p7

My victim VM is at 12.4-RELEASE right now. I'll upgrade it to 13.1-RELEASE first. Then do the next step and see if it also wants to merge config files or not.

alright

i'll be here

i'm going to rollback that jail again

unless somebody else has something to try before I do

rwp, you might be interested to know this doesn't happen to all jails

i was able to update another jail on the same release, and it had to download a lot more patches. on the order of 14,000+

My VM is almost finished upgrading to 13.1R.

5f4de655-1d4f-4f61-8ab0-6c1937d7745b successfully upgraded from 13.1-RELEASE-p7 to 13.2-RELEASE!

i'm not really sure what it's getting hung up at

the only difference is that i stopped the jail, cloned it, and performed the upgrade while the jail was still off

It just finished. It took all of that time to get from 12.4R to 13.1R. Starting the upgrade to 13.2R now.

On the major upgrade the only files that needed merged were passwd and sshd_config and I recall one other trivial one.

yeah that's normal

/etc/passwd will need a look over if it 'looks reasonable' when the 13.1 -> 13.2 upgrade is working properly

and maybe ssd_config's banner

FWIW my VM is a 32-bit system so 13.1-RELEASE-p3 is the latest instead of -p6 and that confused me momentarily.

There are always comments which have changed in these files.

you'll know it's not working properly if it drops you into vi to resolve a merge conflict with /etc/mtree/

"Fetching 4945 patches....." so heading into the next half hour of downloads and upgrades before I know anything more.

No updates needed to update system to 13.2-RELEASE-p0.

No updates are available to install.

Run '/tmp/tmpbf5toe6a fetch' first

can i ignore the run /tmp/whatever ?

I am shocked by the message to run /tmp/tmpbf5toe6a fetch as that just seems really odd.

it looks similar to freebsd-update but it's not

the sha256sums are different

You could diff the scripts.

that's output from iocage update --pkgs 5f4de655-1d4f-4f61-8ab0-6c1937d7745b

Ah... It's an iocage thing. That would be less shocking.

it looks like a slightly less complete version of freebsd-update with some defaults changed like not creating a zfs bootenv

some minor string changes like:

< See freebsd.org/platforms for more info.

> See freebsd.org/platforms/index.html for more info.

Title: Platforms | The FreeBSD Project

Title: Platforms | The FreeBSD Project

rwp, do you know if turning the jail off while doing the upgrade had any effect?

Do I know if stopping a jail kills all processes running in it? I don't know. I would guess that it would.

i'll test that

no i mean having the jail be off before running iocage upgrade or running iocage upgrade while the jail is on

The merge part of the upgrade from 13.1R to 13.2R just started and completed wanting me to merge only /etd/passwd and /etc/ssh/sshd_config and no other files.

lucky you

The merge of /etc/passwd seemed confused about one thing. I normally run root using /usr/local/bin/bash but I always change that to /bin/tcsh before major upgrades. After having locked myself out on a previous upgrade I learned that lesson. Ports might be broken.

I did that and then it remembered my bash choice and forgot that it should be tcsh. And logging in it is back to bash again. Fortunately not locked out.

I had run chsh to make the change so it should have updated the passwd database okay. Strange events.

um

I miss the days before they made passwd into a database. There is miniscule performance difference and the flat files are simple.

it shouldn't be editing /etc/passwd manually or letting you do it - that's a generated file

only /etc/master.passwd should be edited

> "I had run chsh to make the change"... That should be updating the database.

yes, chsh will update /etc/passwd, /etc/pwd.db etc. from master.passwd

Rebooted and the VM is full up to date both base and ports and it was quite simple. If somewhat tedious to wait for it and watch it spin.

And that's all I know.

I am thinking sfox that something in that particular jail smashed the systems idea of what were the core files and therefore it now thinks all of them are modified and must be merged.

But it is late here so I must bid everyone a Good Night! Good luck sfox!

rwp one more question before you head off

did you upgrade sequentially through every point release or did you go directly from 12.x to 13.2?

thanks for helping me

I upgraded from 12.4 to 13.1R first. Then from 13.1R to 13.2R.

thanks

goodnight

I could only share your pain. I had no real help. But I am thinking that whatever is used to know if files have changes has been smashed in your jail and therefore it now thinks all files have changed.

I think you should create a fresh jail for 13.1R and then upgrade it to 13.2R and will find that it works okay.

Then can create another 13.1R and then compare the files in the two jails and figure out what is different.

When the jail is stopped the files are just files on disk. I think you should be able to fix up the broken jail and then it will be fine.

+ Removing jail process FAILED:

jail: ioc-web3: mount.fdescfs: /zroot/iocage/jails/web3/root/dev/fd: not a mount point

Or you could just create a new jail for your whatever task you are using it for and then not worry about it.

started back up though

thanks for the advice rwp

hopefully not too many of my jails are screwed up like this

The /dev/fd is the file descriptor mount point. Not normally mounted by default. But available to be enabled if needed.

I don't normally enable it. But I know others that do.

rwp by the way i upgraded production web3 while it was on

it worked

besides that fdescfs error, even package upgrades worked

and that weird tmp file

Woot! And then you know that it isn't a global problem with all of your systems but just something broken on that one jail.

well several jails are like this

but not all of them

AH!

you know what!? I think the jails having problems with the upgrade where jails migrating from ezjail to iocage

also freebsd-update IDS complains a lot

about what

Title: 270722 – net/syncthing: Update to 1.23.4

when will this be available to latest 13.1 ?

meena, almost every system file being different then the expected hash

is it possible to just ... reinstall a freebsd release ontop of an existing install to fix it?

Hmm, I may have to use this irc client instead heh.

jbo; if you want closure, i tried booting that laptop again with a liveusb, and it turns out not even the network card registers, so its a yikes on my count and i remember why i never tried this earlie

are there any good youtube channels for more technical/development stuff about freebsd? the most popular ones seem to be just user stuff (configs, running programs, etc)

youtube is mainly 101 stuff even for linux videos

you'd probably be better off with books for what your looking for

trev: i reckon the devsumits, so far as they are taped, would be the place to go looking for that

tho me bsdcon talks tend to get very technical, too

sfox: yeah probably need to read something. just i was looking for stuff to passively listen to while doing tedious work

something like fasterthanlime, but less entertaining, given that it's supposed to be background chatter

he writes very long blogs

I do wonder if people design their talks with YouTube in mind these days: "this is a one hour talk that you can put on in the background, tuning in and out.", "this is a twenty minute deep dive, that has you at the edge of your seat", "this is a ten minute wtf demo", "this is a five minute funny one, you'll probably watch it on the loo, don't try

to laugh too hard, but your colleagues know what you're doing anyway"

their incentivized to only brush over topics with what they can fit on a powepoint

yes most likely, since society seems to be diverging in that direction

maybe i should be looking for something too technical, while passively listening, but the channels i found are extremely dull (no offense if any freebsd youtubers are idling)

i watch on 2.0x speed to see if there's anything of value

when an app in ports is available as compiled app in repo ?

Title: 270892 – net/syncthing: update to v1.22.1 at least

as "binary package"

eoli3n: probably a week

ok, it was done 6 days ago : bugs.freebsd.org/bugzilla/show_bug.cgi?id=270722

Title: 270722 – net/syncthing: Update to 1.23.4

eoli3n: build it from ports if you're in a hurry

imm_: i don't want to install ports in that jail, i'll wait

eoli3n: build a package on host and install the binary in the jail

waiting is an option, i'll take it :)

let's do it for the science

imm_: can I produce a .pkg ?

eoli3n: yes

where is this documented ?

Title: pkg-create(8)

see pkg(7)

package target

meena: line numbers now disabled on small displays

imm_: pkg only creates a pkg for an installed package

the package is installed in the jail

not on the host

Title: Solved - Where does "make package" put the .pkg file? | The FreeBSD Forums

from what I understand, installing a port creates a pkg file and installs it. I think ports also have a target to create the package directly, but I don't know

Title: FreeBSD / src / d95fbf4 / riscv: save the thread pointer in both modes - FreshBSD

whewhoo!

if I installed a package in v1 with pkg install file.pkg, when v2 will be in the binary repository, will it get automatically the update ?

upgrade of syncthing in the jails : done

joneum, hi

nginx-1.24.0,3 need pkg-plist fix: /usr/ports/www/nginx/work/stage/usr/local/libexec/nginx/ngx_mail_ssl_ct_module.so:No such file or directory

same for libexec/nginx/ngx_stream_geoip2_module.so and libexec/nginx/ngx_stream_ssl_ct_module.so

build without MAIL options

oh, with all MAIL options

and without GEOIP

eoli3n: sorry, I meant ports(7)

manual pages are nice.

eoli3n: you should get the update for the next binary update, yes

Has anyone made an ISO of the FreeBSD CLI screenreader? Want to take it for a test drive on a spare box.

imm_: ok thanks

screenreader?

You mean watch(8)/snp(4)?

Anyone happen to know a workaround for the issue with not being able to run "pkg boostrap -f" on FreeBSD 13.1-RELEASE-p7?

Well, the command can be run but it just hangs and does nothing. I've tried several mirrors with no change.

CrtxReavr: indeed; apparently there is a patch that provides a console screenreader using eSpeak.

Title: ⚙ D35776 vtspeakd(1): Initial version of console speaker daemon.

sektor, `watch -W <pty>` is a total BOFH tool.

sektor: that would be nice

CrtxReavr: oh wow, that's...interesting.

hrm, no Unicode Support in vtspeakd

i'm suprised that command still works

the linux kernel doesn't have the frameworks to support it anymore

sektor: I think OpenBSD might have something, i know that Stefan Sperling built an OpenBSD with/for a blind friend

no, i misremembered: bsdcan.org/2019/schedule/events/1047.en.html

Title: BSDCan2019: Building an accessible OpenBSD laptop

Maurice runs -current!

ah, there is a video! youtube.com/watch?v=Ma_Y1hVmK8o

Title: Building an accessible OpenBSD laptop - YouTube

what do you suppose "mode" is under dev.pcm.[unit].mode ?

not in the man pages

what device type and what fbsd version?

and what value are you seeing there?

usb audio device - CURRENT. i see 3 for one and 7 for another

i'm trying to sort out my gaming headset's channels and see if there's a way to properly map them, if that is even possible. it's really bugging me that i can't control the volume on the system (only physically on the headset itself)

"mode (1=mixer, 2=play, 4=rec. The values are OR'ed if more than one"

so the device with mode 3 is playback only, the one with mode 7 is record and playback

RhodiumToad: ok makes sense. thanks. where did you find that?

source code

also sysctl -d would have shown it

oh wow

sigh...so much to read, so little time

I guess the issue is how the driver is interpreting the usb device descriptors and creating pcm devices for them

what's the output of usbconfig -d ... dump_all_desc for the device?

RhodiumToad: brb one second. need to reboot since i unplugged it and now the fd is busted

RhodiumToad: it says i can't parse any of the /dev/dsp devices

with usbconfig -d

nevermind. i'm dumb

the ugen* device

assuming it shows as one device rather than two?

yes, one device

sorry, i am slow wiht getting the text out of tmux...i don't have all my stuff setup yet

s'ok

you can paste to termbin.com using netcat, if it helps

somecommand | nc termbin.com 9999

RhodiumToad: that was helpful, thanks! termbin.com/2vb7

not sure how helpful this output is though

by usb standards that's a pretty complex device

yeah it is

interface 0 is the "chat" channel. and interface 3 is the "game" one. "chat" should have a rec out too

interface 0 has no endpoints, so while it says "chat", that's not (afaik) what it is

interfaces 1 alt 1 and 2 alt 1 are unidirectional sound channels, one each way

interface 4 seems to have two alternatives, both output-only channels

interface 5 is the buttons

my usb knowledge is a bit sketchy, so I'm not sure on the details

yeah that seems all fine. i did a little programming against one of the interfaces to control the settings

the way it created the sinks in mixer is just strange

let me check my usb sound dongle to compare

there is a crazy pulse audio patch for this device : pulseaudio/pulseaudio 814419c

Title: alsa-mixer: add support for Steelseries Arctis 7 headset · pulseaudio/pulseaudio@814419c · GitHub

my sound dongle has a basically similar descriptor layout just without the extra device

Heh, dongle.

It's a fun word.

when i set the "game" to the default unit it sounds fine but i can't control anything. when i set the "chat" to default, it's obviously shit sounding but i can control both volumes

meen

meena: interesting stuff.

trev: you might want to try building a kernel with SND_DEBUG and USB_DEBUG enabled

i can do that. what should i be looking for though?

dmesg will have more info?

when you say you can control both volumes, do you mean independently?

dmesg will have more info (you might need to turn on bootverbose) and also there are sysctls to set debug verbosity

should allow some visibility into the probe/attach process

RhodiumToad: yes independently. it's very weird. also the "game" has the mic for some reason, when the "chat" device should

sektor: I think it was really good!

anyone here with wifi who can share their ifconfig -a output?

meena: what's the hypervisor for #269874?

yuripv: kvm / qemu with Q35 profile

ok, was wondering if it's bhyve (it seems to set the pci class correctly to 0x10 for rng at least)

yuripv: i haven't tested linux, because uhh… i don't really care about linux lol

-a, not just wlan\d?

meena: it's good comparison :)

V_PauAmma_V: just all, cuz I care about all, but wifi in this specific case

I'm writing up an email to -net@ and i wanna be sure I'm not talking complete nonsense

if lspci lists the same 0x00ff00 for the class, well, you'll have to live with "old" as the class :D

yuripv: I can actually probably check very easily

given that this is lxd / lxc, it's a lot easier than checking FreeBSD (as of yet…)

meena: please do

meena, bsd.to/pd2j

Title: dpaste/pd2j (Plain Text)

V_PauAmma_V: maybe this is a silly question, but: why does iwm0 not show up in the ifconfig listing?

it does (parent interface: iwm0)

there's a sysctl variable somewhere with the recognized low-level wifi devices

wlan is a generic shim that sits between the hardware driver and the network stack proper

If you mean as an interface of its own and not just the "parent interface" mention, I don't know.

the idea is that "iwm0" isn't actually a network interface, just a driver that if_wlan can use as a backend

(so that all of the generic wifi work can be done in just one place)

thanks. this just leaves me with more questions about infiniband

meena: for lspci, use -nn, the class would be in []