how can I debug things in /etc/make.conf? I need to examine the value of a variable. @${ECHO_CMD} "hi" either in a Makefile or in /etc/make.conf does not work ("Need an operator

it seems to work for basically everybody else on the Internet

nor does '@${ECHO_MSG} "hello world"' as is documented here: docs.freebsd.org/en/books/porters-handbook/book

Title: FreeBSD Porter's Handbook | FreeBSD Documentation Portal

(§5.11 Slave Ports and MASTERDIR). it doesn't work in either make.conf or a Makefile.

moreover, I have a thing that is failing, how do I convince 'make' to show *ALL* of the commands it is executing? currently, I only get a very unhelpful error message referencing a file that appears over 600 times on my system

the first ~10 google results for "freebsd port build verbosity" are not helpful at all

most results seem to refer to the verbosity of the build system (eg. cmake) but not freebsd make itself

try this one something else, few days with somethimes error but it looks good websearchengine.net/?q=freebsd+port+build+verbosity

fury__: make -V ECHO_MSG ?

the resulting output is just "echo\n". no work is performed.

tuxy: unfortunately none of those results are at all helpful or ones that do not appear on Google searches.

I'm also having trouble overriding things in make.conf. if I manually edit this port's Makefile to include 'USE_PYTHON=distutils allflavors', things work as I (somewhat) expect - except that it's looking for an unknown file that doesn't appear. If I edit make.conf thusly:

.if ${.CURDIR:M*devel/boost-python-libs}

USE_PYTHON=distutils allflavors

.endif

it's as if make never sees this setting and doesn't respect it

and the change has no effect at all

(this is why I originally asked how to debug make.conf - I wanted to ensure my `.if` was being respected, but, it seems I'm doing everything correctly)

fury__: -V just prints the variable, it doesn't do any work

great. I still get "Need an operator" every time I try to use @{ECHO_MSG}.

how are you using it?

paste your shiz, man

eg.: @{ECHO_MSG} anywhere in a Makefile - at the top of it, at the bottom of it, in other blocks that use @{ECHO_CMD}, or anywhere in /etc/make.conf

i don't think you can do that, it needs to be in a target

ok. in a target has no effect, even in `pre-build:` - and I need to print to the console from make.conf.

that problem is secondary, though: I need to know which of the 627 'setup.py' files on my system it can't find

for that, I need make to output every command it is doing, which should be incredibly easy and very clearly documented

pre-build:

^I@${ECHO_MSG} "wtf man" works fine for me

what I'm trying to do is `make FLAVOR=py311` in '/usr/ports/devel/boost-python-libs', because that python version is evidently not supported in the ports tree yet (although it should build fine). secondarily, I need to be able to modify make.conf to override USE_PYTHON to be "distutils allflavors", but just for this port.

lol, py311 doesn't even build on my system.

well "^I@${ECHO_MSG}" is very different from "^@${ECHO_MSG}", the latter of which is what is specified in the documentation, so, perhaps that's a documentation bug.

however, whitespace seems to be significant; I do get an echo if I use \t instead of spaces, so, that's progress

yes, makefiles use tabs. it is signficant

^I = \t

so it sounds like that was on problem

~20-year FreeBSD user here, did not know that hah

unless you've hacked on a makefile, it's probable you wouldn't be aware of it

(emacs editor colors wrong spaces in makefiles bright pink as an indicator)

ok, so, I can now see that make is respecting my make.conf changes, and while the USE_PYTHON variable is set by the time the `pre-build` step happens, it's not being respected or used unless it's in the actual Makefile and the changes to make.conf have no effect (but it works fine if I alter the Makefile)

and again, that problem is secondary: I need to know which of the 600+ 'setup.py' files it is not finding

re: hacking makefiles, I've just done a fresh `portsnap extract`, so, nothing is hacked.

hacked as in 'editing'.

understood. again, fresh `portsnap extract`, nothing edited.

aside from some currently commented-out lines.

i don't know flavors/ USE_PYTHON well enough to determine that

if you want to pastebin your build output for the setup.py mystry i can look

I think all you need to know is USE_PYTHON=distutils allflavors in the Makefile makes it able to detect my python version

ok

it may be a waste of your time, but, all that is needed to reproduce this is USE_PYTHON=distutils allflavors in /usr/ports/devel/boost-python-libs/Makefile and `make FLAVOR=py311`

the ultimate goal is to produce a make.conf change that can fix this issue

Python 3.10 has been out since October of 2021, and there's evidently no support for it in this port (or many ports, it seems), and, my team and I need to move on from 3.9 because the advantages of 3.10+ are significant. we have one dependency that relies on devel/boost-python-libs, and so, I need to be able to build it with FLAVOR=py310 (and FLAVOR=py311) from ports

i can't even build python311 so i don't know if i can duplicate that problem

well, either way, I don't think it's going to be helpful for you to run the same command as me and get the same output. I appreciate the input, but I really just need to know how to get `make` to print *EVERY* command it issues.

that should point me in the right direction

you might chcek out make -n or -N

you might have to manually recurse a bit but that might get you started

both produce the exact same output - a single line, and the output is not helpful at all.

Any of the 162108 could be invoking 'setup.py'

er, any of the 162108 files in this directory

is that single line something like cd /usr/ports/devel/boost-python-libs && make CONFIG_DONE_PY39-BOOST-LIBS=1 /usr/ports/devel/boost-python-libs/work-py39/.stage_done.boost-libs._usr_local

grep -R setup.py * shows a README.txt

so run make -n on that make

"Don't know how to make usr/ports/devel/boost-python-libs/work-py311...', or, the same exact output as naked -n if that argument is omitted.

sorry, there is a slash in front 'usr', lest you think it's a spelling mistake.

oh so your flavor is altering the workdir and that confuses it

there has *GOT* to be a way to get make to display every command it issues.

does that directory exist?

it does not, but, keep in mind that there are no Makefiles in this port.

...

so, I will not be able to eg. invoke make from the work directory

instead, I need to know the exact commands that make is issuing

make extract should create it? does it not?

if you're unfamiliar with boost, it uses its own build system. it's quite nice and is better than cmake or make or gmake or whatever if you know how to use it, but, it doesn't use any Makefiles.

I have a work directory, but again, no makefiles.

i've used it; i wasn't super keen on it

but, the ports 'make' does something

after extract

sure. it extracts it to work-py311, applies patches, etc., but, no Makefile is ever available

cd ..

(er, woops lol)

you know, this would be a bit easier if you just pasted some output rather than transcribing or just describing what's happening

what output are you looking for? it's extremely minimal and doesn't contain much more than I've said

except for prompts, etc.

if there was more than one line of output, I'd maybe paste it if it was relevant

===> Building for py311-cython-0.29.33 Unable to find pgen, not compiling formal grammar.

running build

so I'm just back to "I very simply need to know what commands make is invoking, as that will point me in the right direction." This should be incredibly easy.

I consider it a bug if it is either not easy, or not documented.

it appears it's neither

well obviously

but it sounds like you want boost build system to print out commands

perhaps, but, I'll start with make outputting the commands

which is clearly the first thing in the chain

and i told you how to do that

and neither of us know whether it's make or the boost build system issuing that command (it's very likely the former, but again, we'll see once I can see what make is doing)

and neither of us know whether it's make or the boost build system issuing that command (it's very likely the former, but again, we'll see once I can see what make is doing)

you did, but, `make -n` and `make -N` produce commands that do not work

then your build is broken

and reference files that do not exist by the time make exits

nothing on my system is "broken".

I frankly take offense to that.

I've just built this library on Windows of all places. it works fine.

the current issue is that I cannot get `make` to display the commands it is issuing.

I've been doing this a long time; were I able to see the commands being issued I'd be able to see what's going on. I cannot. That is frankly a bug.

ok, if it's printing commands that don't work, it's broken

it's printing commands that reference files that don't exist.

perhaps it deletes them after the make invocation (that should surely be another bug)?

i don't think so

go into your port; make extract; make -n and pastebin the result

Title: dpaste/OVm2 (Plain Code)

as I said earlier, basically. make -n prints a command which does not work and references something that doesn't exist.

side note, `\t@${ECHO_MSG} "wtf man"` does not work *at all* anymore, it only worked the once. it has absolutely no effect.

(even after another fresh portsnap fetch extract)

the very worst experience you can have as a developer is something where you think something can happen, but the result is absolutely no effect at all. another bug.

it's a complicated system

fury__: if it was in pre-config, there's only one pre-config until you make clean

this is in pre-build, and, I already ran make clean

who knows if that portion of the Makefile is executing, though, becuase I cannot PRINT TO THE CONSOLE during it.

this is unbelievably frustrating

what should be an extremely simple issue to resolve is hampered by complete lack of documentation

i suspect you can't just toss around changes to USE_PYTHON like you're doing

well, perhaps, but again, if I could print to the console to debug, and if I could get make to output the commands it was be emitting, and if those things were documented, this would be a complete non-issue and we wouldn't be here. but here we are.

i'm looking at things printed to the console.

and, furthermore, USE_PYTHON is the documented way to tell make which flavors are availble (docs.freebsd.org/en/books/porters-handbook/book §5.11 Slave Ports and MASTERDIR).

Title: FreeBSD Porter's Handbook | FreeBSD Documentation Portal

ffs if I could even get it to invoke `python3.11 -v` as the python interpreter I could maybe get somewhere, but, I can't

`PYTHON_CMD+= -v` in Makefile produces "py311-boost-libs-1.81.0 depends on executable: -v - not found", another really bizarre bug

ALWAYS tell your user what they're doing wrong, and if you don't, you're the problem not the user

fury__: thing is, make has been around for 30 years

that's fine, document it

the search query "FreeBSD Makefile terminal output" on any search engine should have instrutions for this in the first result.FreeBSD"

woops. "in the first result"

ok, then write some documentation

26000 got made with this or lesser documentation

complaining on irc to a bunch of people who have mostly zoned out is not productive

I just wrote about 14KiB worth of documentation on how to build this on Windows. I'd love to. But, chicken-and-egg: I can't document this, as it's completely undocumented. I require two VERY simple things: printing debug messages from Makefiles (or make.conf), and, showing the commands that make generates. These two things are fundamental, and if they are missing, it is a bug.

I know. I'm venting, I appreciate your willingness to help and for being with me on this journey. I might have to make a mailing list post. I appreciate your patience and your company on this ride.

`make FLAVOR=py311` should have absolutely completed my journey here. It didn't, and, here we are.

if you're not super familar with the ports system from th businss end, this is probably not a great starter port to be modifying

it's a huge black box that reads your makefile and makes a port, unless you're quite familar with 12000 lines of port makefiles

i'm tring to bump one port a minor version and also am stuck

Hi

I have installed gtk-mixer and I think my USB mic is on /dev/mixer9 but I cannot speak in a meeting (ie mic not working). The sound (out) is working. I can see youtube etc. How can I fix this?

confirm the level of the mic and that it's not muted

rtprio  imgur.com/W9jhe6e.png   pulseaudio --version

pulseaudio 14.2

rtprio imgur.com/N0QwDEn.png

rtprio do I need to do kldload snd_hda  ref docs.freebsd.org/en/books/handbook/multimedia/#sound-setup

Title: Chapter 8. Multimedia | FreeBSD Documentation Portal

you want to look on the capture tab

rtprio surprisingly no capture tab in mixer9

when I unplug usb mic, mixer9 goes away

comes back as option when I plug in

so I guess thats the usb mic

rtprio that usb mic only has "playback" option.

maybe that is why its not recording/capturing

rtprio but    cat /dev/sndstat

Installed devices:

...

pcm5: <Realtek ALCS1200A (Rear Analog 5.1/2.0)> (play/rec) default

...

pcm8: <USB audio> (rec)

pcm9: <USB audio> (play/rec)

No devices installed from userspace.

pcm9 here says play/rec

but /dev/mixer9 has no rec in gtk-mixer (I am running as non-root user)

rtprio any ideas?

are you certain it's mixer9?

when I unplug usb mic, mixer9 goes away

try just `mixer`

rtprio termbin.com/0r9mq

mixer -f /dev/mixer9 mic.recsrc=+

rtprio mixer: unknown device: mic.recsrc=+

usage: mixer [-f device] [-s | -S] [dev [+|-][voll[:[+

..

 devices: vol, pcm

mixer -f /dev/mixer9 -s

rtprio /dev/mixer9 -s

vol 100:100 pcm 100:100

it doesn't look like that particular mixer supports rcording

is thre a mixer8?

am.. but I was using that mic on linux before.

there is mixer 8 too

mixer -f /dev/mixer8 -s

mixer -f /dev/mixer8 -s

mic 100:100

that's the one to use for recording then

rtprio but it stays there if I unplug, also no sound goes through . HOw can I check?

mixer -f /dev/mixer8 mic.recsrc=+

that should enable the mic

rtprio mixer: unknown device: mic.recsrc=+

usage: mixer ..

 devices: mic

i don't know man, i don't use sound; read `man mixer` again

rtprio I found the issue. It is using mic of mixer5. How can I undo it and use mic of mixer8/9?

strange that this isn't working either `mixer -f /dev/mixer8 +=rec mic`

fury__: how about you read makes documentation and find the -d flag that lets you debug the build as much as make can let you, you probably want loud mode as a first step

How do I change my recording device?

I am trying to update a port but receive "[00:00:01] Error: Nonexistent origin net/sfwbar". Any ideas please?

mixer -f /dev/mixer8 =rec mic

 what is wrong with this ?

Title: Internal microphone setup question | The FreeBSD Forums

simply not working for me

nor this tiny doc man.freebsd.org/cgi/man.cgi?mixer(8)

Title: mixer(8)

these commands work for mixer5 but not mixer8/9

so driver issue?

now I have the error "[00:00:00] Error: No such ports tree development". Any ideas please?

elgrande: where are you seeing that one

did you poudriere -c -p development … before?

and to your nonexistent origin from before: are you really trying to update it? I can't see net/sfwbar in my copy of the portstree

One or more devices has experienced an unrecoverable error. An

 attempt was made to correct the error. Applications are unaffected.

action: Determine if the device needs to be replaced, and clear the error

errors: No known data errors

zpool status -v ^

Putittali: what's smartctl say?

meena says  termbin.com/bgwf

the only thing i see is unsafe shutdowns

yes and 6% use in 150 hrs

that's not much use. maybe it's bored?

ok :)

# mount /dev/nvd0p3 /home/user1/nvme0/

 mount /dev/nvme0p3 /home/user1/nvme0/ -> mount: /dev/nvme0p3: No such file or directory

 # ls /dev/ |grep p3   -> ada1p3 dsp3.0 nvd0p3 nvd1p3 nvd1p3.eli

mount: /dev/nvd0p3: No such file or directory

gpart^

I'm reminded of those MFBF graphs: they're bathtub shaped, so 150 hours of operation for the first instance of failures to show sounds reasonable

How do I know what FS is it?

I installed freshports.org/sysutils/fusefs-lkl  but cannot mount a partition that may be LUKs or BTRFS. What can I do?

Title: FreshPorts -- sysutils/fusefs-lkl: Full-featured Linux BTRFS, Ext4, XFS as a FUSE module

is the fuse kernelmodul loaded?

how do I know?

kldstat

nimaje grep fuse don't show anything

nimaje how do I load it?

 kldload fusefs-lkl

kldload: can't load fusefs-lkl: No such file or directory

kldload fusefs.ko

 kldload fusefs.ko loaded it

why is cirrus-ci.com/task/6091282020302848 running on 13.1-RELEASE despite the pull request is against main?

Title: Cirrus CI

elgrande: why not?

elgrande: the kernel build *of* (your patch of) main is running *on* 13.1-RELEASE

it's nice to be able to build CURRENT on a supported RELEASE, don't you think?

can anyone loop me in on what the latest re: openssl is around QUIC support, the v1->v3 shift, and kTLS? my understanding so far is openssl v1.1.1 is EOL this year, v3 breaks a lot of things, and OpenSSL is implementing their own QUIC that'll take half a year more. however, only OpenSSL has kTLS

is anyone using one of the alternative libs like boringssl, wolfssl, libressl, etc? do I keep kTLS with security/openssl-quictls?

domlaut: FreeBSD CURRENT has cgit.freebsd.org/src/log/?h=vendor/openssl-3.0

Title: src - FreeBSD source tree

oh, so freebsd is migrating to openssl 3

that, rather than 3.1, i reckon, is likely going to be what's will go into 14.0-RELEASE

we also have bearssl: cgit.freebsd.org/src/log/?h=vendor/bearssl

Title: src - FreeBSD source tree

yeah, on the QUIC thing, I don't think anything <3.4 will be all that useful to me (that's when they expect it 'done' IIRC)

not sure if that's got anything to do with kTLS, but it's used in the loader

loader?

the thing that loads the kernel

Title: loader(8)

ah, thanks. interesting choice

well, it's absolutely tiny, so it's probably the most sensible choice (given that we also have to consider licenses;)

yuripv: I had expected that 14 is build on 14, but I did not know anything about it.

s/build/built/

meena: so it is build in a jail, right?

elgrande: i don't think so github.com/freebsd/freebsd-src/blob/main/.cirrus.yml

Title: freebsd-src/.cirrus.yml at main · freebsd/freebsd-src · GitHub

meena: so it is crosscompiled for each platform under 13amd64 or under a separate 13 platform for each arch?

meena: yeah, I'm working with streaming stuff over the internet so my set of requirements is different (stuff like needing DTLS, etc). unfortunately due to how building the system works is I can only link against one ssl library for ingesting/broadcasting/serving static files/all of it :-)

so it seems what I'll have to do is depend on openssl 3 and then do static linking of boringssl/quictls/libressl just for quic

elgrande: not for each arch… only for the Tier 1 archs: amd64 & aarch64, but with different compilers

which compiler is used on aarch64?

read the thing, my brain is bleargh right now

I don't want to :)

well, i can barely read even when my brain isn't bleargh

what is the "correct" way to have neovim to be used instead of vim ( nvim instead of vim )

if I don't have a vim installed

on debian there is something called /etc/alternatives which manages links to alternative binaries

in freebsd I just symlink nvim to vim and case closed or is there alternative to debian alternatives?

...

merp: we don't really have such a system in place

so, yes, just add a symlink

and I did, thanks

Does anyone of you use a Yubikey with Firefox?

Hi

GELI encrypts disks while OpenZFS encrypts datasets  . Which one I need? I heard there are issues with latter gist.github.com/rincebrain/622ee4991732774037ff44c6768085ab.   What is the best way to a) formate and then b) encrypte a zfs in linux? is it any different than a *bsd ?

Latiute: geli is full disk encryption, like linux's LUKS. it's bsd only. zfs can run on top of the geli container.

zfs has encryption, but some metadata isn't encrypted, just the data.

that's very new

linux is openzfs, too

yeah

Demosthenex elgrande so geli is stable enough? also at installation time, what technique does zfs uses to encrypte drive?   geli?

No.

geli has nothing to do with ZFS

I see

I have heard that there any more than one encrytption methods for zfs

how is the drive encrypted then?

Latiute: geli's fine. and if you use the installer and select encryption, it will setup zfs on a geli partition

symetric

it is FDE - full disk encryption, dealing with your disk blocks

you can install ZFS or UFS over it

Demosthenex what does the freebsd installer do/use at installation time encryption?

it uses geli

I see

so symtric fde

can I use geli in linux? Actually I am backing up data from btrfs drive to zfs drive (yet formated and to be used on freebsd )

No.

*installer* can not offer anything else than geli, because we have not teached boot loader to grok zfs encryption yet.

you can  use zfs in linux

Latiute: installer does root zfs on geli

angry_vincent no means i cannot use geli on linux

Demosthenex elgrande I see

Latiute: that's right, geli cannot be used with linux

manual installation does not provide encryption

no means you cannot

Latiute: linux uses LUKS

so my only stable option is to not  use encryption

that would work on both linux and bsd

ya, bsd has no support for lUks

iirc

perhaps. you need similar zfs versions.

i wouldn't bet on sharing a disk between bsd and linux ;]

Demosthenex well just a one timer data transfer

sure. consider making your final storage geli on zfs, and just unencrypted zfs for the transfer

so in summary, zfs dataset encryption is NOT be used as its not so stable yet?

alternatively, look into zfs send/recv

it's new, and both bsd and linux have different support for it

which one is new and both l...?

like angry_vincent said, you can't boot with encrypted zfs for root because the bootloader doesn't understand zfs encryption. but you may for a data disk

Demosthenex rest udnerstood

Latiute: both. it's a new feature, and depends on what version of zfs bsd and linux are synced to

either way, for stability, i'd use LUKS+zfs on linux, or geli+zfs on bsd for now

both geli and zfs enc dataset?

No

without zfs encryption

one let me wrap:  consider making your final storage geli on zfs, and just unencrypted zfs for the transfer <-- understood

for your data migration, look at zfs send/recv

either luks + zfs be used or geli+zfs. But not zfs dataset encrytion. Correct?

yep

Demosthenex first I have to foramte a drive. Then copy data into that froma a btfs drive

then maybe zfs.send() somewhere

oh, well that's right, if its btrfs to zfs, send/recv won't help

.   What is the best way to a) formate and then b) encrypte a zfs in linux? is it any different than a *bsd ?

sorry

ignore b)

zfs encryption can be used with data pool. That is, pool without boot disk(s).

most folks say setup a partition table and then add zfs

i say use the disk raw, no partitions

tsoome yes but we just discussed that zfs encryption of dataset is not so stable

Demosthenex me too. I want to use raw disk too. do you know how exactly to do it?

Demosthenex I know its 1-2 commands onlinux but I am very confused from different articles from internet

have you done that before yourself?

Latiute: on bsd make a geli on the block device for the drive, ie: /dev/da1

or have any idea?

just a sec

you made it?

Title: dpaste/mRn9n (Python)

i may have missed the geli attach call though

Demosthenex well that is ok. Will be useful at when I am on bsd but for now I have to read btrfs and bsd can't do that with luks encrypted btrfs drive. For that I have to use linux. In linux I can't have geli. So need simple zfs drive

it's geli init each disk, geli attach to open them, then zpool create

so how to formate a drive and create zfs in it

that paste

was when i setup my new server

i installed my OS to 2 ssd's, that was done by the installer

the paste is where i built a 4 disk RAIDZ ZFS on GELI

without partitions

what is the maximum compression comand instead of zpool create -O compress=lz4 -O atime=off -f zdata raidz da2.eli da3.eli da4.eli da5.eli

there are newer compressions, but i'm using an older one

zstd

is thsi ok  zpool create -O compress=zstd ssd1 /dev/sda

Demosthenex ^

if you want max compression, you probably want to read about zstd;)

elgrande VimDiesel I had a lot of issue sand work stuck. So don't want ot make mistakes after ready docs/blogs as an inexperienced person

tsoome yes, I have been using that in btrfs too. its nice

tsoome is the command I gave you correct and ok?

zpool create -O compress=zstd ssd1 /dev/sda

yea, I mean, compression=zstd will not give you maximum

oh it will not?then what will?

Latiute: honestly, use some compression, but don't worry about "max"

Demosthenex its for backup

need max

Title: ZFS - thoughts on zfs & compression | The FreeBSD Forums

tsoome ^

 sudo zpool create -O compress=zstd ssd1 /dev/sda

invalid vdev specification

use '-f' to override the following errors:

see zfsprops(7), it does explain

tsoome zst-1000 then :)   You can specify the zstd level by using the

       value zstd-N, where N is an integer from 1 (fastest) to 19 (best

       compression ratio).  zstd is equivalent to zstd-3.

       Faster speeds at the cost of the compression ratio can be requested by

       setting a negative zstd level.  This is done using zstd-fast-N, where N

       is an integer in [1-9,10,20,30,...,100,500,1000] which maps to a nega‐

       tive zstd level.  The lower the level the faster the compression - 1000

       provides the fastest compression

Latiute: does your backup tool also do compression?

disk compression + backup compression = double compress = wasted space

lz4 and zstd both can detect if incoming data can be compressed.

I thought zstd couldn't.

but indeed, you do not want to have duplicate work

tsoome: yes, it's not as bad as it used to be

Demosthenex my tool is rsync

what zstd level be used if 70% compresion needed?

Latiute: You're asking about Linux in one channel, and FreeBSD in another?

thats assuming your data is compressable:D

Compression ratio isn't something you can decide when using in-line compression.

compression depends on the data type.

You get ratio you get, as a function of the CPUtime you're willing to spend.

w/o hardware accel (ie: lz4 on an LTO drive), just use some compression.

There's only hardware acceleration for gzip via QAT, iirc.

compression is trading cpu cycles to throughput. the higher compression options will eat more cpu (and memory). So, you kind of need to test it out.

yes I know. text vs videos

I know all that. Was generally speaking

Luckily, even with recordsize<1M, the memory requirements aren't as punitive as they can get on really large sets of data.

nimaje: make -d -A produces what looks like over 30k lines of output, so not human-readable, and does not tell me the specific commands it's executing

fury__: as I said, you probably want loud mode as a first step (not letting make tell you really everything it can tell you)

in any case I appreciate it, more than I had before at least

Using python3 shutil library seems to not be using the right PATH to find installed binaries is there something else I need to make it work as expected? Code snippet I am trying here: paste.rs/IYB.py3

where do you use it? what do you expect? what do you see? what is PATH there?

Using it in my home dir. Running `which rustc` on the cli outputs the PATH as I would expect

the path it outputs is /home/user/.cargo/bin/rustc

so which rustc and python -c "print(__import__('shutil').which('rustc'))" differ?

Yes the output on the python script as you wrote it is None

hm and python -c "print(__import__('os').environ.get('PATH', None))" matches PATH in your shell?

Looks like it: /home/user/.cargo/bin:/sbin:/bin:/usr/sbin:/usr/bin:/usr/local/sbin:/usr/local/bin:/home/user/bin

Yeah they are the same outputs

ok and python -c $'import os\nimport os.path\nprint(os.path.exists("/home/user/.cargo/bin/rustc"), os.access("/home/user/.cargo/bin/rustc", os.F_OK | os.X_OK))' does that output True True ?

True True

hm, then I don't see why it shouldn't work; you use py3.9? and that isn't a dir?

Python 3.9.16 specifically. What do you mean isn't a dir?

it is a normal file and not a directory?

Yeah not a dir. Python dir? Sorry still not sure what you mean

Interestingly I am finding things like py-whichcraft in ports (freshports.org/devel/py-whichcraft) but that seems odd to need.

And more googling found this issue from 2019: bugs.python.org/issue35755 but is notes as resolved

Title: Issue 35755: On Unix, shutil.which() and subprocess no longer look for the executable in the current directory if PATH environment variable is not set - Python tracker

^ but note it was fixed with a patch?

the last check in shutil,_access_check would be not os.path.isdir(fn)

that doesn't apply to you as PATH is set

ok, no idea why it doesn't work for you, the code seems to be correct, but should find /home/user/.cargo/bin/rustc for you, maybe step thru shutil.which('rustc')

ok I will try and run it with pdb

stepping through seems to identify the rustc bin at least in some capacity. paste.rs/q5w

This is weird

Though it shows usr/bin/rustc instead of .cargo/bin/rustc

Yeah stepping through it guessing for rustc in a few places but not the right places

If the proper .cargo/bin/rustc path is in my PATH I wonder why it never actually looks there

Found this as well: python/cpython #79936 so I will keep tinkering. Thanks for your help nimaje

Title: On Unix, shutil.which() and subprocess no longer look for the executable in the current directory if PATH environment variable is not set · Issue #79936 · python/cpython · GitHub

79936 – [MAINTAINER] irc/riece: Mark BROKEN on ia64 bugs.freebsd.org/bugzilla/show_bug.cgi?id=79936

that also says 'current directory' and 'if PATH environment variable is not set' in the title at least

Yeah just google scrambling for anything that seems sort of related

Hey. I am just trying out FreeBSD from Linux. I have a Thinkpad P14S AMD (Gen1) Laptop. I have tried everything as suggested by the handbook and everyonline but I can't get sound to work.

This is the output of sndstat

Tried with chromium / mpv. No audio

I really want to ditch Linux and this is the only hurdle I am facing

acommonman: You probably need to change the default sound output. Try this command: sysctl hw.snd.default_unit=3

dumbbell, let me try. thanks friend

acommonman: You may need to restart yout browser after changing the default, I don't remember if it will re-open the correct device.

omg that worked

dumbbell, you are a saviour

thanks man

I will put it to sysctl.conf

acommonman: Cool :) According to the content of /dev/sndstat, the first three outputs are audio over HDMI, corresponding to possibly up-to-three HDMI connectors

dumbbell, just one little problem. the headphones are not switching though. it still plays from the laptop speakers. Do I need something?

acommonman: My guess is that plugging in a headphone in the jack connector won't do what you expect (i.e. mute speakers and send the sound to the headpĥone)

let me test it

Right ok, we came to the same diagnostic :)

dumbbell, ya that worked

perfect

acommonman: Do you mean that when you plug in your headphones, the speakers are muted and the sound goes to the headphones?

dumbbell, no. it was not switching automatically. i had to manually sysctl to the 4th device

how do people get auto switching to work? pulse?

acommonman: No, you need to set another variable to configure the device

Oh o

*ok

acommonman: Can you share the output of "sysctl dev.hdac" please?

sure. wait

acommonman: Can you run "sysctl dev.hdac.1.pindump=1", look for lines with "hdaa1" in /var/log/messages, and share those lines?

Sure

acommonman: Afk for diner, I'll come back after :)

No problem. no rush. I will be around. Meanwhile I will just paste it for you to see

I've never quite figured out nid switching, all I know is that I've been lucky that the speakers mute when I insert a pair of headphones on my ThinkPad T480s.

I'd love to learn though, so will be keeping an eye on this conversation :)

debdrup, hello. ya well. I guess its a hit or miss thing. I mean if nothing works, I might just create aliases with "speaker" "headphone" or something. Dirty hack, but I don't mind it

Worth leaving the mess linux is becoming for good.

Another thing I need to figure out. How to get GUI applications inside jails working with wayland

Documentation is weak as most users still use X11

acommonman: I'm back, looking at the pindump you shared

dumbbell, welcome back

acommonman, debdrup: HDA (snd_hda(4) allows the user to configure the routing of inputs and outputs to tune things like "all the outputs form a single 5.1 sound system" or "those two outputs are a single system with a headphone which overrides the speakers when plugged in"

Ah ok. dumbbell thanks. i will look into it

acommonman: In the pindump you shared, we see the two outputs you are interested in: the NID 20 (speakers, with caps = out) and NID 33 (headphones, with caps = OUT)

This is confirmed by the pcm3 and pcm4 inits:

pcm3: <Realtek ALC257 (Internal Analog Speaker)> at nid 20 on hdaa1

pcm4: <Realtek ALC257 (Right Analog Headphones)> at nid 33 on hdaa1

That's right

acommonman: You need to add the following two lines to your /boot/loader.conf:

hint.hdaa.1.nid20.config="as=1" # Speakers

hint.hdaa.1.nid33.config="as=1 seq=15" # Headphones

Oh wow. Let me try that

hdaa.1, because hdaa.0 corresponds to your audio over HDMI outputs (the kernel finds this one first)

Ya I got that. we are dealing with the analog stuff here

"as=1" on both settings means that both output will now be in the same group (and thus will form a single pcm device instead of two before)

"seq=15" is to indicate which one takes precedence (when it is active, i.e. something is plugged in)

Ah. So that's what. Thanks for the detailed logical explanation.

You will need to reboot by the way, I didn't give the instruction to reconfigure the device at runtime

yes I understand. Let me try this out

Linux kernel as a much larger "database" of "computer/motherboard/laptop identifier => use this HDA configuration by default" and does a better job at extending it

FreeBSD kernel has way less of these entries and end users often have to configure that themselves (which is some kind of black magic unfortunately, even on Linux...)

Ya I know. I am not at all expecting linux level out of the box experience. Linux is linux and it does what it is designed to do. I just want a more simple system. If most of the things I need work well with a bit of manual intervension, its fine. I dont' care. The lasted two days I used FreeBSD, i have absolutely loved its simplicity, the separation of the system and pkg (specially the /etc gosh linux /etc is such a mess)

acommonman: Did it work?

acommonman: That's cool you love FreeBSD so far :)

I am going to try it out now. I was kinda occupied watching this video lunduke.substack.com/p/convincing-a-linux-guy-to-use-freebsd

Title: Convincing a Linux guy to use FreeBSD - by Bryan Lunduke

O let me try it

Ok rebooting. bbs

dumbbell, it worked

wow

Great!

dumbbell, are you like a freebsd dev? you should definitely update the handbook if you can

acommonman: I am. Everything is documented in the snd_hda(4) manpage, but the discoverability is... difficult to say the least :)

Let me check. I want to improve my rtfm skills for freebsd. I do love the documentation though. I really hate asking stuffs that I can find out

ok found it. yes, its definitely a little difficult for the average person. But the documentation is wonderful. It at least mentions everything

Yes, but the problem is how you find this man page from "I plug my headphone and this dumb system doesn't mute my speaker for god sake!"

dumbbell, urmm. man page will be integrated with chatgpt soon? lmao. jokes

Yes I know the system should handle it and spare the OS from this

like what the hell

All of this is what makes Linux so damn bloated with so much code

wait till it's all flatpacks. You have not met the bloat...

Hahaha, that would be cool :) But at that point, ChatGPT should just figure out the configuration itself and update /boot/loader.conf

But you prove an age old saying. BSD users understand Linux "or I would say in depth low level knowlege" more than linux

jarebear6expepjo, I am a massive flatpak user. haha.. I use Gentoo Linux. So I kinda gave in to flats last year to save compiling chromium

But unlike Ubuntu + apt and snapts, Gentoo at least didn't force it on me

acommonman: Something I never figured out yet was how to configure snd_hda(4) to support a headset (with an integrated mic) plugged into that single jack connector

Well I suppose you can always add disks to your storage arrays :D

jarebear6expepjo, btrfs man. btrfs is the shiz. lol joking. Sorry. I know we got zfs here

dumbbell, I do have wired mobile headsets like that.

I don't need to solve that problem.

But ya, knowing it would be good

acommonman: The headset might work for you with a bit of luck

ya I will tinker with it. Although I have bigger things to fix. Like getting gui applications inside jails to work with wayland. I use sway.

Like, if I can get brave browser and chrome to work for netflix i am done with Linux like for good

I did succeed in installing jailed linux and everything. but gtk stuff didn't work with my wayland

Hello, I looked at my system and nothing from dumbbell is configured on my system, but, the switch between laptop sound system and the plus of a jack head work since… the insttallation of FreeBSD on that machine 2017 (i guess)

Anyway. That aside. I do want FreeBSD and other BSD projects to stay the way they are. They don't need to fall for those "automated" / "unified" trap. As long we have each other as a community. Like you solved my problem today. I don't think we need weird workarounds

dumbbell, thanks a lot to you

plug*

Lovis_IX: Either your computer HDA routing is correct ouf of the box, or FreeBSD knows how to configure your device. Ideally, it should be the case of all laptops, but that's not the case unfortunately.

FreeBSD is such a complete system that I was shocked to find the "backlight" program being part of the system. Under linux I had to use xbacklight for that

acommonman: I never tried to jail my graphical applications, so can't help with this.

acommonman: Hope you'll have a fun journey on FreeBSD :)

dumbbell, its ok mate. you have helped me enough. I will research it on my own or someone else might get me through it later..........

dumbbell, and thanks so much

You're welcome!

Beautiful system and such a skilled community

And goodness gracious. I thought the days of OSS' Analog Audio quality over ALSA were over. I am listening to losst mp3 and I can still find it better. May be a placebo or something, but I don't know

*lossy

Its just listening to Laura Branigan's Self Control with chromium + youtube under freebsd with headphones and I absolutely love the quality

I always thought DACs decide good audio. Its interesting how OSS / ALSA still fairs.

Anyone, turning this question to everyone. I have a question. I heard its a bad idea to mix using FreeBSD with ports and binary pkgs but people still do it. Which one is still pragmatic. Is it better to use FreeBSD predominantly from source with ocassional binaries to save time, or is it ok to use freebsd with predominant binary packages with pkg and using ports only if a package needs customizing

I am asking this as a Gentoo user who doesn't mind compiling world

*anyway

I would mostly use FreeBSD from ports and use the occassional binary from pkg for things like chromium / libreoffice. Is it a good idea?

acommonman: I personally use packages for everything, except when I want to change an option

acommonman: It's possible to kind of mix packages and compiled ports in a clean way. You need to use Poudriere, the tool used to compile the official packages. It's easy to setup locally. This way, it compiles packages with the options you want. Then you simply install the produced package(s)

Poudriere is something I need to look at

tcpip stack should only use ARP request if the dest IP is on same subnet as src IP, otherwise forward to gateway right?

Hi, wanted to discuss some storage architecture

Don't know if few would be interested to read

What is usually done by people these days? pool of drive acting as one big drive or raid configuration? in either case, we can add/remove drives dynamically and it won't matter? what if disk encryption via GELI is there for each drive?

first I thought to have 1tb disk for database, 1tb disk for web app files, (lots of pictures, data), 1tb disk for OS, 1tb disk for another database.   Sometimes I think why not just combine all those drives to make one via raid0 (to save money, size (combined usages is less and easy with one big volume). And I have to take backups anyway to

another drive even if I use raid 10 or raid6. Then sometimes i think to use raid 6 with double parity. All else same. I know nothing about zfs vs mdraid. Its a long journey but wanted to have an informed decision so weeks and months ending in frustration does not happen

btw, i have 2 nvme 1tb each and 2 ssds 1tb each for now. Another 2tb 5" hdd and 1tb hdd for backups maybeeach database will be at least 1tb, web data may grow to 2tb

 and then all above will start multipling each 6 months

there are many options but the point is which strategy in brief i use?

-------

short question: problem : save storage  (best use of it) without affecting performance

dumbbell: oh, I was hoping that my dock, which has a couple of 3.5mm outputs on it, would register so that find a way to use it.

The USB ports also don't appear to work correctly (though this may have changed on 14-CURRENT as there were a fair number of usb changes since I last updated), but for some strange reason the RJ45 connection does work correctly, as does the power and displayport+hdmi connections.

The old T420 I had had a working dock audio setup, but that appears as a completely separate snd_hda(4) device.

But it's entirely possible that that was thanks to acpi_ibm(4)

I'll give what you wrote another re-read when I'm less sleepy.

following up on my openssl woes from earlier today, I decided to go with libressl and forego ktls for the time being

whoever came up with DEFAULT_VERSIONS for make.conf -- very cool

Title: ports - FreeBSD ports tree

It's a lot more recent than I thought it'd be.

yup. last time I compiled software was prior to that change. by about 2 or 3 years. :-)

is there a way to sync ipfw tables across multiple hosts ?

Unfortunately not, no.

Believe me, I wish we had stateful synchronization like pfsync provides.

I wish so too :)

when is nethack 3.6.7 going to be available in the ports tree?

3.6.7 fixes a serious security fix.. thats why im asking.. it says so on the website

fixes/includes