In case anyone is curious, on my php-gnupg adventure ... it's a bug in gpgme, and potentially the PHP module as well.

> so.. we can't just download and install enlightenment is what you're saying

kindred: there are packages, <freshports.org/x11-wm/enlightenment/#packages>

Title: FreshPorts -- x11-wm/enlightenment: Artistic and fast X window manager

has freebsd been working on a replacement for openzfs because even though its not GPL licensed software, it is still weak copyleft even just on the files

...no?

it's still CDDL, no?

yes

it was CDDL before freebsd switched upstreams to use openzfs

replacing it is not a feasible project

So. I found a potential bug in php*, combined with php*-pecl-gnupg.

Namely, the latter can never work unless I manually modify `php-fpm.d/www.conf` to add /usr/local/bin to $PATH.

(it needs the `gpg` executable & friends)

is it explicitly setting $PATH, or is it inheriting it from somewhere else?

By default (as www.conf is currently packaged), $PATH is empty.

Which means that unless I manually modify www.conf to add /usr/local/bin to $PATH, php*-pecl-gnupg simply won't work.

(and it took me *way too long* to figure that out due to [non-FreeBSD-specific] bugs in both gpgme *and* php-gnupg --- the former errors out when it shouldn't, and the latter silently ignores it when it shouldn't)

If you want the full story: community.passbolt.com/t/the-openpg…-used-to-encrypt-in-freebsd/7800/15

Title: "The OpenPGP key can not be used to encrypt." in FreeBSD - #15 by darkuranium - Installation Issues - Passbolt community forum

$PATH is actually set to an empty string?

(linked reply is the most relevant one)

RhodiumToad: no, it's simply not set (I believe). PHP-FPM's default behaviour appears to be to clear its environment.

procstat -e can usually display the environment of a process

I'm guessing Linux distros override that to make `php-gnupg` work or something?

Yeah, I just did a <?php echo getenv('PATH') ?>. Was an empty string.

Not sure how I'd get procstat of an existing process. But ANYWAY.

Most likely just missing, considering. I could check, but it's 2:30 AM and I kinda wanna go to sleep >_<

procstat -e somepid

Ah. Thanks.

The usual daemon `PATH=/sbin:/bin:/usr/sbin:/usr/bin PWD=/ HOME=/ RC_PID=62165 BLOCKSIZE=K`.

But www.conf does *not* set `clear_env = no`, and default is `yes`.

execvp() uses _PATH_DEFPATH if PATH is not set in the environment

So php-fpm clears it.

I had to do `clear_env = no` (for other reasons), plus:

`env[PATH] = /usr/local/sbin:/usr/local/bin:......`

and _PATH_DEFPATH usually includes /usr/local/bin

The latter is the most important one.

Yeah, I think it just inherits it from the "root" `php-fpm` process otherwise. Which is above.

I guess ultimately from rc?

usually the /usr/local/* dirs go at the end

Well, either way, point is, they're missing.

what freebsd version?

12.4-RELEASE. Oldish, I know.

rc does set PATH to /sbin:/bin:/usr/sbin:/usr/bin overriding daemon's login class

Yup.

I guess maybe this could be fixed in php-fpm's rc script? To unset PATH?

... though it still wouldn't be a full solution unless `clear_env = no` *or* `env[PATH] = ...`.

(but a simple `env[PATH] = $PATH` would work then)

clear_env can't be clearing $PATH

if it were, it would have worked

Well, no. It clears PHP's `getenv()`.

I think it basically clears it for the spawned subprocesses.

no, I mean it can't be clearing the environment for subprocesses

can't be clearing PATH specifically for those, that is

Aaah. Right, because then they'd get _PATH_DEFPATH?

Maybe it *is* setting it to an empty string, then.

dunno. maybe try invoking env as a spawned subprocess and see what that says.

Maybe if you remind me tomorrow, I *really* should go sleep >_<

(sorry!)

Good night ^^

Thanks for the help.

lets say i create a zfs device pool with one device at /dev/da0. if i move to to another drive controller, how to i replace the device path in the pool with the new device name of /dev/ad2 and preserve the data?

I don't think you need to

it'll just locate it based on the metadata when the zfs service is running?

CCFL_Man: with ZFS, when you give a device to a pool, the device gains four copies of a vdev label. The label includes the pool_guid number. Labelling will persist beyond future actions such as removal, detachment, and destruction.

<openzfs.github.io/openzfs-docs/man/8/zpool-remove.8.html>

Title: zpool-remove.8 — OpenZFS documentation

<openzfs.github.io/openzfs-docs/man/8/zpool-detach.8.html>

Title: zpool-detach.8 — OpenZFS documentation

<openzfs.github.io/openzfs-docs/man/8/zpool-destroy.8.html>

Title: zpool-destroy.8 — OpenZFS documentation

those labels are a menace, as shown by recent incidents

Clearance of labels is typically performed with zpool-labelclear(8) <openzfs.github.io/openzfs-docs/man/8/zpool-labelclear.8.html>.

Title: zpool-labelclear.8 — OpenZFS documentation

CCFL_Man: you might like to check for the presence of labels on the first of the devices with which you tested, the device from which you 'moved', for example:

zdb -l /dev/da0

This assumes that you used a whole disk (not recommended) for the test.

RhodiumToad: no.

clearly yes, given the number of people suffering breakage

Each of the four copies is a lifesaver, in cases where (for example) removal, detachment, or destruction was mistaken.

doesn't help those people whose systems were not bootable due to stray labels

Edge cases.

I'm intentionally dispassionate, because I have in mind an enhancement to documentation.

when using bhyve is performance any better by running VNC inside the VM and connecting to it, or using the EFI console and connecting to that via VNC?

Saves you extra software

ooray, openss 3.0.9 landed.

crb_: both are software based framebuffers, so they're not gonna be fast no matter what - and any performance difference is likely to be small

debdrup, Are you excited by the news alone or are you running 14 with the OpenSSL 3 installed?

if your guest os is windows, rdp is faster

parv: mostly that re@ can start working on 14.0-RELEASE again

debdrup, Aye

the real thing i'm excited for with openssl 3.x is that ossl(4) can be extended, and that eventually aesni(4) can be retired

ossl(4) can be extended to do what aesni(4) does now, but for a much broader set of architectures (both assembly-optimized and SIMD-optimized versions)

During 2-3 recent attempts I "Stuck on ERROR-tried-to-rebuild-during-make-install,..." during installworld (log: lists.freebsd.org/archives/freebsd-current/2023-June/003859.html ; /etc/{make,src}*: lists.freebsd.org/archives/freebsd-current/2023-June/003860.html . That is due to c 2017 bugs.freebsd.org/bugzilla/show_bug.cgi?id=212877#c21 . How do I proceed?

Title: Stuck on ERROR-tried-to-rebuild-during-make-install, from 1cebc9298cf to b98fbf3781, or from 9fbeeb6e3831 to dfa1982352ee

Per one message of "/tmp/install.PjBzUKhv0k/sh: cc: not found" -- indeed there was no "cc" in that temp directory

Alternatively, what would I need to do to force the build/installation of whatever missing bits the message points to?

parv: make clean?

meena, Yeah, just started that process. (I thought had tried that once by accident|mistake as I had missed to update the object & ccache directories for one of the attempts.)

when shit really gets bad, i make clean, and then blow away the obj dir

Will do

do you have any fancy options set?

i haven't seen any catastrophic failures lately

Ooh oh, do I?! See lists.freebsd.org/archives/freebsd-current/2023-June/003860.html mail for the options

Title: Re: Stuck on ERROR-tried-to-rebuild-during-make-install, from 1cebc9298cf to b98fbf3781, or from 9fbeeb6e3831 to dfa1982352ee

I have the memory of someone else around Feb-Mar also having the issue after llvm|clang was imported but have failed to find that message so far.

Few days ago did not occur to me to do world building from zero🤦

anyway ...

hello. anyone running Homeassistant on freebsd? after upgrading to python3.11 and creating a new jail, i'm getting attributeerror module 'urllib3.util' has no attribute 'protocol_tls'. i'm not sure if it could be related to Freebsd's openssl flavor?

well, if you have VM space available, openssl 3 just got merged into -current, could be an easy way to check that. I'm currently running kyua on a fresh -current build for similar reasons now

though as a disclaimer, I don't run homeassistant and I'm not totally sure if it's openssl related. It would be my hunch though.

how about installing the openssl3 port?

is there even an openssl3 in the ports tree? I would think that would conflict with the 1.1 in the base system, I thought that was the whole point of the openssl3 merge

hm i just told pip to upgrade urllib3, it did but then it complained about a conflict with botocore

fonks, pkg search says there's openssl30 and openssl31.

i think i can live without that. it's a python library for AWS

HA seems happy now

V_PauAmma_V, yeah, I'm eating my words now, I should have run a find on /usr/ports before I said anything.

hjf_, python related issues are entirely out of my depth. I avoid python for that reason :P pip likes to make package management on both the underlying system and the python env a hellscape

fonks: well in order of mess you have CPAN, PIP and NPM

cpan being the worst

npm's approach is nice because all packages are kept within each app's directory... but that makes each app liek 800MB of dependencies

yeah... no thanks. Storage may be cheap but that's a bit rediculous

reminds me a lot of JS dependency cycles. At least python doesn't have everyone and their dog making their own frameworks

i was developing an app on "nestjs" the other day, wondering what's really the point of using it

nestjs is a metaframework of several other JS libraries

it has very verbose dependency injection for no real benefit

as long as your app is small everything is good. then you get into circular dependencies and everything breaks. it's like it forces you not to have "cross dependencies" in your project

when i do simple JS APIs i KISS and just install express for http, jwt for auth, typeorm for DB, winston for logging, and typescript because javascript sucks

I'm a php guy. Processing is the server's job, imo. Clients are and should be treated as dumb document viewers as much as possible, in my mind.

JS, on top of being a mess, is not really something that belongs in a browser. But I digress.

i hate doing frontend but "browser is a dumb document viewer" doesn't really cut it anymore

for work i had this very specific requirement which was that, it's a call center with 1000s of people working. they have this "real time dashboard" that show what everyone is doing at every moment

kyua found a few failures on 1400092, but I also forgot it was running as root which can be iffy. I'll shoot those over to the mailing list once an unprivileged run finishes

(handling a call, waiting, training, lunch, break, etc)

hjf_, yes, that kind of application is understandable, that's why I said "as much as possible." :P

basically the dashboard had a timer that refresed it every few seconds... we "upgraded" it to a websocket to stream what everyone is doing

server is happy that it's not being hit with refresh, DB is even happier, and clients are happy because they can see things with at most 5s delay instead of 2 minutes lol

I just detest having entire application environments in a browser, it's really not good for performance or security to have that. Blinkenlights and status pages, understandable. Video streaming? Sure. But stuff like google docs, or the insane web version of Office? That's getting a bit iffy, in my mind.

yeah that's too much

but writing CRUDs in React is nice imo.

and having the same language betwen FE and BE has a big advantage: 1) TS Types can be shared across both projects (declare a DTO and use it in both. change the backend? the FE build breaks, you know you have a problem), and 2) validation logic is as easily shared between FE and BE

form validation in FE is a nice thing but often you have to repeat it for BE and that sucks. with a shared validation module, hnnggggghh

I recently had to do some stuff in ArcGIS, and was introduced to ArcGIS online... and promptly returned to ArcGIS Pro. Performance was garbage, had weird naming restrictions, and was implemented at least 90% in the frontend

so my browser was doing all of the (barely-threaded, because JS) heavy lifting.

my next project for work is to migrate the company's IA core

AI

the core is a python service copy pasted from examples from 10 years ago that's just a neural network

it's a pain to deploy it, it's an unmaintainable mess (because i hate python but had to maintain it), etc

trying to see if tensorflow.js can be used in the backend with the real C tensorflow

"When you import TensorFlow.js from this package, you get a module that's accelerated by the TensorFlow C binary and runs on the CPU. TensorFlow on the CPU uses hardware acceleration to optimize linear algebra computation." CRIES IN HAPPY

Hopefully that goes well o7. I'm not really versed in AI, I know vaguely how it works but never dealt with the nitty-gritty. Hopefully all that training data doesn't have to be converted or shuffled, if it's been running 10 years

what is the best way to get linux apps installed on freebsd? should I got with ubuntu based system? docs.freebsd.org/en/books/handbook/linuxemu

Title: Chapter 11. Linux Binary Compatibility | FreeBSD Documentation Portal

fonks: it's basically a huge matrix in a file. hopefully JS can import it

Beladona, honestly? I usually go through a process of (1) look for it in native packages/ports, (2) try to find source for the linux version, compile it natively, and run it natively, and (3) if all else fails try to run the linux binary with the linux emu. It should be built into the default kernel IIRC, so you should just be able to run the binary directly.

a problem i actually had with tensorflow is that (LOL) the VM i was assigned didn't have the CPU instructions to run the TF binaries.

this company squeezes every cent of the hardware and software they buy

whenever we have to deploy something new they keep asking if "we can make it compatible with SQL 2008 because we have some licenses we can reuse"

hjf_, yikes, I thought it was in some DB

Keras models (typically created via the Python API) may be saved in one of several formats. The "whole model" format can be converted to TensorFlow.js Layers format, which can be loaded directly into TensorFlow.js for inference or for further training.

yassss

i think we use keras in the python version

Beladona, you will probably find that most of your "linux software" exists as native packages. Try 'pkg install <thing>' to learn more, or browse /usr/ports for names of packages (or install them from source in /usr/ports directly). Copying over linux binaries is sort of a last resort, but can and does work.

No new permissions-unrelated failures running kyua as a regular joe. Most seem to be related to dns limits, one utf8 locale failure, and a single lock-order failure in mixer.

fonks brave browser with no ads do not exist. having issues lib conflicts with obs-studio and kdenlive

Can you elaborate on the lib conflicts? I've not used Brave, but on first glance it appears to be open source, you may need to build from source to get a native version running

fonks imgur.com/4f1chnm.png

obs-studio

fonks kdenlive says it will remove postgres

Beladona, sorry for the delay, was filing a bug report. Have you been installing everything from pkg or from ports? Do you have a mixture of software going on? Pkg is really good about not nuking a system like that unless something is very wrong

fonks this is the vendor / lib issue

you can try to install it too and you will see. Install kdenlive, obs-studio, postgresql, neovim fonks

I need all those, what should I do? fonks

Please allow me time to respond. I'm doing other things in the background, too, you know. :) As far as installing all of those, I want to know first how you installed those.

fonks sure

# chroot /compat/ubuntu /bin/bash

chroot: /bin/bash: No such file or directory

fonks for installing, just use pkg

what to do of this error?

Beladona, what are you doing exactly? the freebsd linux compatability works like wine in a sense, you load the linux binaries and the kernel's linux compat shuffles around the syscalls to match freebsd calls. You don't need a chroot of a linux environment and that in fact will not work

please see docs.freebsd.org/en/books/handbook/linuxemu

Title: Chapter 11. Linux Binary Compatibility | FreeBSD Documentation Portal

you don't need to, and should not, install an entire linux system as a subtree

fonks I already installed ubuntu via `debootstrap jammy /compat/ubuntu`

right, which you can do, but what is your goal?

fonks I need calmness and peace of ubuntu's package manager who can install brave, obs-studio, kdenlive and other linux apps for me.

fonks how can I run `apt install brave` now?

By running ubuntu? I'm not sure why you're trying to essentially run linux but with the freebsd kernel

You can install Brave natively by building from source, I should mention, in the freebsd environment with the freebsd kernel

I have freebsd as base os, with that I need mentioned apps. those create problem on freebsd. so I installed ubuntu on it. Those apps are big and simple binaries will have problem. as said I need peace of mind for ubuntu's package maanger to auto manage dependancies

fonks how can I run `apt install brave` now?

belatedly, fonks, I've had similar lua dependency issues. For example, you can't have mpv and minetest packages installed at the same time.

You also can't have KDE or Mumble installed at the same time, last time I tried to install KDE.

Beladona, I don't think that's what the linux chroot system is for. I've only ever used it to build linux libraries for linux binaries that needed them. If you're trying to use apt, your best bet is to use ubuntu, either in VM or on the bare metal.

Like I said before, the order of preference from most to least is (1) pkg, (2) /usr/ports, (3) source package compiled natively, (4) source compiled for linux, (5) precompiled linux binary

xtile, lua dependencies are strange. I've noticed that before with minetest. I wouldn't think kdenlive would want to obliterate postgres though, unless there's some other shenanigans going on

fonks no.. see docs.freebsd.org/en/books/handbook/linuxemu/#linuxemu-debootstrap

Title: Chapter 11. Linux Binary Compatibility | FreeBSD Documentation Portal

What a horrible idea.

jkc which one?

Beladona: What you're trying to do.

jkc well then how to make the installation of kdenlive, obs-studio, postgresql, neovim simple?

Beladona, what FreeBSD version are you running? on 14-CURRENT all of those will install without conflict from pkg and ports

13.2

Which version of postgres do you want?

Because on 13.2, pkg install kdenlive obs-studio postgresql15-server neovim works just fine.

jkc it doesnt

 imgur.com/4f1chnm.png

Beladona, what else have you done with this system? It sounds like package management has been pretty badly broken by something, and from what you've said so far it's not clear what

pkg install kdenlive obs-studio postgresql13-server neovim

fonks the snap clearly says it is a bad dependancy conflict

The incompatibility is in the version of postgres you're trying to install.

jkc I have a serious database v15.3... why install 13?

the pg db is already running

That's your problem. The version if pgsql you installed is what's causing the conflict.

Just use "vi". Fixed. :)

well why?.. the version of neovim is also causing conflict with kdenlive or whatever

jkc if those packages are part of ports, then why conflict?

Beladona: I just told you the actual conflict. You can also switch your pkg source from quarterly to latest and pgsql 15 works just fine, along with the rest.

Beladona, conflicts come from how the packages are built. The default configuration will inevitably not work in certain combinations or circumstances. For that you can file bugs and follow -latest, as jkc said, or build from /usr/ports with a custom configuration.

<joke>or you can forgo package management entirely, just drop binaries everywhere, and cross your fingers</joke>

its already latest  in  /etc/pkg/FreeBSD.conf  as   url: "pkg+http://pkg.FreeBSD.org/${ABI}/latest",

jkc termbin.com/4cus

fonks still its too much time consuming and frustrating

jkc fonks what you said seems correct but apparently not working. How can I change that?

what are you referring to, exactly?

ok, lets do one by one. how to install obs-studio fonks jkc ?

this is taking a lot time as you can see. there is not straight fix. Or am I missing something?

fonks jkc mentioning you again and again? Are you on this or I should not?

Beladona, since IRC is asynchronous people may or may not be monitoring it at any given point in time. Even if they are shown as "online". I don't have answers for you, but out of 600+ connected people someone else other than fonks and jck might ... eventually.

unixman_home ok. waiting for the remaining 637 people then

:D

Glad to see the "I want volunteers to remain available 24/7 to answer questions for me" is still going strong.

debdrup or maybe we can change the subject of IRC chat to "please do not disturb, I will email you when free"

Is there a way for me to ... somehow wipe anything outside of GELI content in a partition? For some wacky reason I have a system that's finding a phantom pool across two partitions, where those partitions are actually encrypted by GELI, with my actual pool therein.

mason, The best and only tool I have found for that so far is wipefs which does not seem available on FreeBSD.

Title: View paste KCAAW

rwp: So, digging more I see ada0p3 showing up in "geom list" twice as a provider, once with Name: gpt/tank0 (desired) and once with 1. Name: gptid/7d9cbbbf-ad51-11ed-b75e-94c691d5946f (source of the phantom pool)

I can probably just wipe the partitions by hand and rebuild, but I'm wondering if there's a more elegant way.

Possibly unrelated, this box came to my attention as it froze during a "freebsd-update" last night.

but, from rescue media, attaching geli, importing the real pool by id, and scrubbing showed no errors

rwp: If you have ideas I'll try them. Otherwise, I guess I'll wipe each half in turn and rebuild.

As a by-the-by I have been dealing with my own system problems of late and have been deep into this on one of my systems too. Everything is failing all around me.

rwp: Same sort of error?

rwp: Do you suspect an idea is percolating? If so I'll hold off on the wipes as that will likely nuke the opportunity for more exploration.

And there's no rush, so I'll go do cooking dailies on my AzerothCore server and keep this open to see if there are any ideas in the meantime.

My problem is that I don't know exactly what problem I am experiencing. But I am not seeing a phantom signature problem as you are seeing.

How would you characterize what's happening to your system(s)?

Also, I'll take that as "okay, wipe away" :P

My problem is related to endless drive resets between HGST Deskstar NAS drives and the controller in a 6 drive raidz2 pool. Which all started happening mostly all at once across several of the drives.

Ew. I don't want to trade problems. =cough=

rwp: Storms? Power surge? Inductive voltage spike? Can you swap out the controller/HBA/whatever to rule out that flaking out?

I can make the story worse. This is an array that was too big to back up.

Yee. Glad it's on ZFS then.

If you can, I'd suspect the HBA first and try to rule that out. Worst case, now you have a spare HBA.

But I really only have about 6TB of data on it, and 5TB of that is definitely on the previous smaller NAS pool okay. But there are some recently acquired audio books and some other things on the new difference that would annoy me to lose.

For my part, I'm 50G into wiping 4.6T, so, whee!

And I'll have to do it again once I've resilvered the new ........ I can just grab the beginning and end of the partition I bet.

I have three Dell surplus LSI HBA SAS/SATA adaptors. Have swapped them around. Have swapped pretty much everything.

Are all the drives same vintage/model?

So then decided to swap motherboards. Moved the drives into a different system. The problem follows the drives. Which tells me it is nothing about the system hardware. The problem is definitely the drives.

That completely sucks. Can you start swapping the drives out? If you do, I'd think about varying the make/model to avoid systemic failure.

Recommendation that seemed crazy to me not that long ago: look at used drives.

Yes. Unfortunately. 6x HGST HDN726060ALE610 6TB SATA drives. Which is always better for a SOHO type of setup to have decoupled drives from different places to make sure failure modes are not similar. But...

As in, if you can get some reliability out of used drives, that cuts back on the existential threat.

That's what I am doing right now! I bought 2x used SAS drives from a vendor. And 2x from a different vendor. And two of those arrived and were BRAND NEW, well, 12 and 18 hours on them respectively. Woot! Won the lottery on those two.

Nice, wow. So, hopefully you start seeing the problem disappear then.

The other two were 30k hours on them as I expected. And have two more in the order pipeline to arrive probably Tuesday.

And now am in the process of copying the data off the previous SATA drives onto the incoming SAS drives. And learning some things as I go here.

nice

For one... If I try to copy from two of the pesky SATA drives at the same time (using ddrescue) then I get endless drive resets and drive errors and it just blows chunks!

Yeah, just do one at a time.

rwp: Oh, ddrescue... Is it not better to let zfs drive the resilver?

But if I do one disk at a time then I avoid all disk errors. I see softreset in the kernel log. But ddrescue does not report any read errors.

In order for zfs zpool to resilver or scrub or replace one must have an array online. Mine has degraded, suspended, and faulted.

Things went from bad to worse pretty quick. Can't get enough "read" out of the drives to assemble an array. And in the current state I don't want to try.

Instead I will make a complete copy of each drive off of the pesky troublesome drives one by one onto the incoming replacement drives one by one.

Then the replacement drives can be worked with to assemble the zfs pool. And if I screw something up then I still have the original data in the original state to work with.

ugh

kk

A lot of work but a solid plan.

For my part, I'm going to potentially save hours here with a little seek= magic to dd.

got 100M off the front, will do roughly the same off the back

I am glad I can make other people's day a little happier because they can go, "Whew! At least I am not that guy!" :-)

haha

mason kane?

I think there's a demotivational poster about that.

Cane sugar?

Title: Mason Kane | Citadel Wiki | Fandom

Is it easy and advisable to upgrade from 13.2 to 14 ?

Beladona: There's also a character Mason Verger from Thomas Harris' Lambs series. BTW, your nick is missing a couple consonants.

mason intentional

Beladona, Perhaps the old saying "if you have to ask then no you shouldn't" applies? But instead one might ask for what reason would you desire to do so while it is still in development?

rwp am. ok typing:

tried to install ubuntu based system? docs.freebsd.org/en/books/handbook/linuxemu

Title: Chapter 11. Linux Binary Compatibility | FreeBSD Documentation Portal

because:

 brave browser with no ads do not exist. having issues lib conflicts with obs-studio and kdenlive

 imgur.com/4f1chnm.png

I had a heck of a time with debootstrap on Linuxulator a while back. Might try it again.

 kdenlive says it will remove postgres

mason how to run `apt install brave` after doing `debootstrap jammy /compat/ubuntu`

Beladona: Did their debootstrap get all the way to completion?

Beladona: Pretty sure you'd just chroot in there if it finished and have you a coherent system, and then use apt or whatever as though it were an actual Ubuntu system.

I just did pkg install debootstrap  and debootstrap jammy /compat/ubuntu in  docs.freebsd.org/en/books/handbook/linuxemu/#linuxemu-debootstrap

Title: Chapter 11. Linux Binary Compatibility | FreeBSD Documentation Portal

And then are you chrooting in?

chroot /compat/ubuntu /bin/bash

chroot: /bin/bash: No such file or directory

Beladona, What is that upgrading from and to? I haven't seen similar problem on my system so far. I am 13.2-RELEASE at the moment. And the quarterly ports neovim is neovim-0.8.3 not 0.9.1 as on your system so I assume you are on the dailies?

Beladona: Is /bin/bash there? What's in /compat/ubuntu/bin?

mason No such file or directory   for both ls

Beladona: from the outer system, please pastbin: ls /compat/ubuntu

rwp I am on latest

pastebin*

I am going to back away because if this is in linuxator then I know nothing. And I my drive number 2 just finished copying after 13h21m runtime and I need to set up drive 3 to start copying.

rwp I am on latest. Not quarterly

rwp: Nice, may it proceed placidly and without excitement.

# ls /compat/ubuntu/

debootstrap/ var/

Beladona: That suggests you might have followed the instructions incorrectly, or that the instructions were incorrect.

what should be the steps?

Beladona: If you debootstrap into /compat/ubuntu then you should see the normal root-of-filesystem things in /compat/ubuntu. You seem to have stuck it all down a level in debootstrap maybe. Can you show us ls /compat/ubuntu/debootstrap?

debootstrap.log debpaths deburis

Beladona: That isn't the expected root goop either. Not sure how you can the debootstrap, but either you pointed it to the wrong place or it broke, which is the option I observed last time I tried.

s/can/ran/

ok doing again then

pkg install debootstrap,     debootstrap jammy /compat/ubuntu

mason do I need to set up mounts in /etc/fstab.

Beladona: The world inside /compat/ubuntu should be independent of the outer system.

ok but this says about  fstab docs.freebsd.org/en/books/handbook/linuxemu/#linuxemu-debootstrap

Title: Chapter 11. Linux Binary Compatibility | FreeBSD Documentation Portal

ok anyway, what to do after `debootstrap jammy /compat/ubuntu`

mason oh, this time it worked

Beladona: It helps when you know what's going on under the hood. :P

Just kidding.

mason pastebin.mozilla.org/J0RPaED3/raw

Beladona: Ah, not sure what you want to do to get valid /dev, /sys, and /proc inside there but you'll need those. The guide you followed should address this. Might be outer-system fstab in that case.

mason ok.. why I get permission denited?   pastebin.mozilla.org/J0RPaED3/raw

Beladona: Read my last line. You need functional mounts as noted.

The /dev/null it wants needs to already exist.

so need to change /etc/fstab in freebsd?

Beladona: You should see what the guide you use recommends to set up /proc, /sys, and /dev inside the compat environment.

mason pastebin.mozilla.org/hyLxs4kK/raw

am fixed. checking

Beladona: Hm, is there no Linux-specific devfs? That's the only thing that'd worry me.

mason I do not understand. I just pasted what was in doc and it worked

but I do not see any much packages. /etc/apt/sources.list is also not present

Beladona: /compat/ubuntu/etc/apt/sources.list?

If there's nothing there or in sources.list.d you'll want to address that, as you won't have any sources of packages until that exists.

Good luck with it, and thanks for the inspiration - I'll give it a try again soon, although with Debian rather than Ubuntu.

its there. but not seen in side the chroot

Back to the cellar to reboot my newly-resurrected system.

just 'main' deb added there.

Beladona: Unclear how it can be there but not from within the chroot.

mason imgur.com/ILNw6te.png

when i do "zfs list" i see "zroot/ROOT/13.2-RELEASE_2023-06-21_063639" ... does freebsd create that when updating?

markmcb: Yeah, it makes snapshots now on freebsd-update

Useful.

that's a nice feature i didn't know about

Beladona: That looks right. Note that you said "/etc/opt" in the left side.

mason inside chroot its /etc/opt

Beladona: Inside the chroot it's definitely not /etc/opt.

bbiab

mason sure pastebin.mozilla.org/8j6H05Bz/raw

markmcb: it's from the new feature added to freebsd-update, to create boot environments when upgrading; you can access old boot environments via the loader

that's so great :)

today's a big day. after a month of trial, i migrated all my core server tasks to freebsd from my linux server. count me amongst the believers, lol

markmcb, Run "bectl list" and then go look at wiki.freebsd.org/BootEnvironments because Boot Environments ROCK! :-)

Title: BootEnvironments - FreeBSD Wiki

boot environments are basically impossible to live without, once you've started using them.

markmcb, Note that beadm is the original tool in ports and bectl is the newer tool in base and both are basically the same within limits of same-ness. See forums.freebsd.org/threads/howto-freebsd-zfs-madness.31662 for history.

Title: HOWTO: FreeBSD ZFS Madness | The FreeBSD Forums

i feel like each week i find a new amazing thing. thanks for highlighting this!

We are all on the journey of discover together. :-)

beadm (and boot environments) originate in OpenSolaris

the beadm tool in ports is a reimplementation of beadm in OpenSolaris

mason it was terrible way to run obs via ubuntu in freebsd. no audio, screen etc. I should have fixed the freebsd compatibility issue instead

I think I should upgrade to v14

It's not -RELEASE yet.

FreeBSD 14.0-CURRENT

I'm not sure that's a good idea.

debdrup whats your version?

am I the only one who can't install obs-studio with neovim and postgres?

pastebin.mozilla.org/rtPSEKu8 - does it make any sense that `fg` in /bin/sh accepts PIDs but not actual job IDs? that's contrary to what the man page says and also to what bash/tcsh do

Title: Mozilla Community Pastebin/rtPSEKu8 (Bash)

Beladona: what's the conflict?

luke8086: %job

meena: that works, interesting. thanks!

it's weird that it's not documented, like it's supposed to be very common folk knowledge

luke8086: it looks like csh(1) documents it, even tho sh(1) doesn't, but it's the same syntax abs behaviour

heh, i've been using linux/unix for 20+ years but mostly bash/zsh and fg <number> always worked :P i guess you never stop learning

indeed in bash/tcsh you can call %job alone without fg. but not in /bin/sh

POSIX strikes again to ruin your fun

mason, As a by-the-by my next disk is copying at a good rate an hour into the copy and no disk resets seen in the log. So this drive is probably fully okay. In the end I may only have three pesky drives out of the six.

rwp: Ah, there were more woes on my end. Bootloader can't find my pool, even though rescue media has no such issue.

Frankly going to just install Debian, as this is something that ought not to be broken and I don't have time for it.

luke8086, If there is no % prefix how could the shell tell the difference between a PID written as a number say 1 and a job number written as say 1? It must have a % there to namespace it to the job number. Always been like that. :-)

mason, I understand. Compromises must be made as a matter of practicality. One must be pragmatic. But nothing is as permanent as a temporary solution!

haha

If I had any hope of actually understanding what's broken I'd dive in, but then things don't get done.

On my problem the curious thing is that this six drive pool has been running since November 2022 without problems until a couple of weeks ago. Scrubs every 35 days. All okay.

rwp: That's the whole monoculture thing. Maybe it's a logic bug in the drives.

rwp, It's one of those 34 days ...

I am thinking that too. Or maybe the drives are all tripping some controller problem. Because it feels like the media of bits is okay but that controllers plural are all freaking out at the same time.

Normally I am one of those who try to get as many different vendors of drives as possible to de-couple all possible coupled failures. But then I had these 6 drives that all tested perfectly...

rwp: The exception that proves the rule.

rwp: indeed, frankly i hadn't known it accepted PIDs, my first thought was that it's a bug that it works with PIDs rather than job ids ;) note that it works without % in bash so perhaps it takes the first match

luke8086, ... or if the number does not belong to a PID?

luke8086, Well look at "kill" which is also part of that set of job control commands. That definitely takes a process id. And it must be built into the shell in order to handle job control processes.

actually it doesn't accepts PIDs in bash

bash has a lot of bash'isms. And most of those bash'isms originated in ksh.

Use zsh. Get both (ba|k)shims

"I didn't really say everything I said." --Yogi Berra

-- nice typo

In the case of job control everything originated in csh first.

meena  imgur.com/4f1chnm.png

also if I install kdenlive, it says postgres15 will be removed

Beladona: I'm a committer, so 14-CURRENT.

Beladona: so the problem is: different versions of luajit

Beladona: is it possible to convince all of those packages to use the same version? if not, which combination conflict? can you run PostgreSQL in a jail?

etc… I'm sure there's more questions that lead to sensible solutions. like, what's different in latest vs quarterly?

hm. luajit-devel is the default luajit version at least in latest

for all three of the above packages?

I mean in default-versions.mk

obs-studio isn't using the default version.

it's using lang/luajit (old 2.0.x release) except on powerpc where it uses luajit-openresty

any good reason why?

that suggests it could use luajit-devel without build problems (if it can build and run on the openresty version, it should be able to handle luajit-devel too)

2.0.5 is known broken on some architectures

luajit is a right mess because although there's been activity again on the 2.1 branch, there's still no release as far as I know

everyone just picks a git commit and uses that

RhodiumToad so in short what is the solution to this?

what do you want to do?

meena no idea. I have to find out. how?

basically, the package builds of obs and neovim are incompatible, so you can't install both in the same system image. You could build one of them from ports

RhodiumToad I want to install obs , pg15 (already have it), neovim and kdenlive (well brave browser too but thats a different story)

what's the issue with kdenlive?

it removes pg

what's the output of the pkg command in that case?

one sec

RhodiumToad kdenlive imgur.com/tcP2QbL.png

gdal??

whats that

gdal: 3.7.0_2

it's a GIS graphics library

ok but what to do here? obs is priority though, actually all are

RhodiumToad and I think this will happen in any freebsd version and release

easiest fix will probably be to build neovim and gdal from ports

how do you know its gdal?

well I don't know that gdal is the only issue for kdenlive, but for whatever reason, it looks like kdenlive is pulling in gdal and gdal depends on postgresql13-client

so the solution would be installing neovim and kdenlive from source?

not kdenlive, but gdal; if you build yourself a gdal having set DEFAULT_VERSIONS+=pgsql=15 then it will depend on pg15 rather than 13

then either build neovim with DEFAULT_VERSIONS+=luajit=luajit or modify the obs port to use luajit-devel

(assuming neovim works with that luajit version, I never tried)

also maybe bug the obs port maintainer to update their luajit dependency to use USES=luajit

ok let me fix obs first. how to build neovim like that. I mean exact command?

are you using latest or quarterly packages?

latest

do you have a recently updated ports tree? have you built anything from ports before?

I just did `pkg update`

RhodiumToad I upgraded from 13.1 to 13.2 few months ago though

RhodiumToad as you are a pg guy, fyi, SELECT version() still says  PostgreSQL 15.3 on amd64-portbld-freebsd13.1, compiled by FreeBSD clang version 13.0.0 (git⊙gc:llvm/llvm-project.git llvmorg-13.0.0-0-gd7b669b3a303), 64-bit

yes, so?

13.0. not 13.2 but not an issue

the 13.0 is the clang version not the freebsd version

ok

so what would be my first step?

" amd64-portbld-freebsd13.1"

first you need an updated latest ports tree

how?

do you have a populated /usr/ports ?

RhodiumToad yes it has dirs

does /usr/ports/.git exist?

no

should I be doing `gitup latest`?

anyone know offhand if portsnap still works, or is it dead now?

RhodiumToad: Updated last month, so it ought to still function.

RhodiumToad: it's more zombie than dead

Oh... Hm.

"IGNORE: portsnap is in base systems and has been removed from 14.0-RELEASE"

from freshports.org/ports-mgmt/portsnap

Title: FreshPorts -- ports-mgmt/portsnap: Provides secure snapshots of the ports directory

should I be doing `gitup latest`?

no

I haven't seen svn purged from base, but, might as well

Not before 13 is EOL.

how to update ports then?

ports tree*

debdrup: why 13? wasn't the last release done in svn 12?

portsnap fetch update should still work then?

Beladona: any of the usual methods.

meena: That's possible, I was remembering it as 13 though.

Title: Chapter 4. Installing Applications: Packages and Ports | FreeBSD Documentation Portal

RhodiumToad I am going to follow docs.freebsd.org/en/books/handbook/…s/#ports-using-installation-methods

Title: Chapter 4. Installing Applications: Packages and Ports | FreeBSD Documentation Portal

Beladona: most reliable would be git, but that would require you to install stuff, and installing stuff seems fraught right now

Why is installing stuff fraught right now?

installing git-lite would be no big deal

doing make install clean

Beladona: you need to update your ports tree first

RhodiumToad ok how to update?

install git or git-lite as a package rather than trying to build it

I already have git

ok, then follow the rest of those instructions to check out the ports tree, but you may want to add --depth 1 to the clone command

I'm wondering if the handbook shouldn't be recommending the tiny flavour of git.

RhodiumToad ok but it says git vulns.

says what?

debdrup: and depth=1

RhodiumToad when I do  make install clean

Beladona: do not try and do 'make' for git

you said you already had it installed

It's only got ftp/curl and textproc/expat as dependencies.

you can unshallow in the background of you really need to, but to get started it would be very useful

yes, git is installed but says vul error

no

start at step 2 of the instructions!

step 1 is just to install git, but you did that already

meena: what's the actual advantage of a shallow tree, other than it takes less space?

ok, I hope this will give me latest. not quarterly . git clone git.FreeBSD.org/ports.git /usr/ports

Title: ports - FreeBSD ports tree

the vulnerability thing is because you didn't update your ports tree yet

use --depth=1 for speed

ok, I hope this will give me latest. not quarterly . `git clone git.FreeBSD.org/ports.git /usr/ports depth=1` *

Title: ports - FreeBSD ports tree

I suppose it's quicker, yeah.

debdrup: less time. if you're trying to get thru a step by step thing, and literally the first step takes half an hour, you night give up on step two

RhodiumToad ok

yes, the default branch corresponds to "latest", each quarterly branch has its own name

meena: right, I suppose I'm displaying my privilege unconciously, as I'm used to working on 1/1Gbps FTTH and a fast NVMe SSD.

Beladona: git clone --depth=1 git.FreeBSD.org/ports.git /usr/ports

Title: ports - FreeBSD ports tree

fatal: destination path '/usr/ports' already exists and is not an empty directory.

deleting dir.

probably better to move it aside but whatever.

interesting things happening

rm -rf /usr/ports/

rm: /usr/ports/: Device busy

Oh, `git-fetch --unshallow` can be run later on - in that case, I'd say there's no reason not to add --depth=1 on the git-clone(1) in the handbook.

pwd

You're likely in the directory you're trying to delete.

I cd ed

where to?

debdrup: our ISP incompetently disabled our SIM so hard, it's no longer connecting after they enabled it. they did that after i asked about IPv6, and they then wondered, who are these people? they've never paid. (because the ISP didn't take money from us lol)

home. RhodiumToad

What about other shells/tmux instances?

meena: sorry, I think I lost the thread of the conversation..

rm didn't worked fully but cloning now

meena: oh, that was a tanned gent, gotcha.

RhodiumToad  git -C /usr/ports pull

Already up to date.

oh, /usr/ports might have been a mount point on a zfs install

Is the ports tree updated now ?

if you did the git clone, then it's updated.

RhodiumToad: yup, it is.

RhodiumToad ok, what next?

next you need to edit /etc/make.conf

add a line DEFAULT_VERSIONS+=luajit=luajit pgsql=15

no I think this is not about pgsql but luajit-devel only

I already explained to you what the problem with pgsql was

ok. doing what you said

btw, this already had MAKE_JOBS_UNSAFE=yes

I think it should be removed

you have pg15 installed, if you want new port builds to respect that then you set it in DEFAULT_VERSIONS

I added DEFAULT_VERSIONS+=luajit=luajit pgsql=15

and removed the unsafe thing

RhodiumToad next?

debdrup: the tanned gent was about slow Internet. the other thing that makes Internet slow in this household is: storms.

Beladona: cd /usr/ports/editors/neovim and do make missing-packages (should output a list)

meena: I used to have a DSL connection (ADSL2+Annex M, specifically) where the copper was so bad it needed pretty heavy attentuation, and despite that every time it'd rain the speeds would drop precipetously.

Beladona: put the output on termbin so we can see it

whew

meena: the ISP just shrugged whenever I'd mention it.

RhodiumToad pastebin.mozilla.org/8CUchL8R/raw

brb 5 mins

uh.

that's lots of missing stuff

it's also completely horked up the flavors

ok. 'make missing' seems likely to be more sane

Beladona: can you do make missing instead and see what that says

RhodiumToad it says devel/cmake-core, security/rhash, devel/ninja, lang/luajit

ok. you'll want to do pkg install cmake-core rhash ninja luajit

oh, you'll need to remove your current luajit-devel, which will remove neovim

true

(can't be helped, you'll have to use something else to edit while we rebuild neovim)

doing that. will reinstall neovim

done

ok, and now 'make missing' is empty?

devel/luv

that depends on luajit too so we'll let it rebuild that

ok

should I install obs-studio now?

(cd /usr/ports/devel/luv; make install)

oh ok

let's finish with neovim first

ok

done

back in editors/neovim, is 'make missing' empty now?

make missing shows nothign now

ok. make install

that should build and install neovim

inside editors/neovim?

yes

ok

it may prompt you for options, you can use the defaults or pick anything reasonable

RhodiumToad we didn't used pkg to install neovim. Is this `building from source` ?

yes

the ports gave us source. and we `make` it

ok understood

its doing it

the port should fetch the source and then compile it

when we did git clone, the ports fetched it already. no?

I wonder why I am getting neovim 0.9 and not 0.8

that fetched the files of the port, i.e. the instructions on what to fetch and how to compile it

that was the default in 13.2

oh ok

the port doesn't contain the actual source of the program, it just says where to fetch it from

I see

neovim installed

ok. now try installing obs-studio from the package

it should no longer conflict on luajit

RhodiumToad awesome ! works

right. next step is to try and figure out why kdenlive tries to pull in gdal; if we can find a way to turn that off, it'll save having to compile gdal

super. what to do next?

give me a sec to try and figure out the dependencies

ok, brb 2 mins as well

why is there no good way to look up the dependency chain between two specified packages

ok, found it. kdenlive -> frei0r-plugins-opencv -> opencv -> gdal

so either building opencv without gdal would probably work, or building gdal with the proper pg version. Probably easiest to do the latter

RhodiumToad ok

but can't afford to uninstall pg as I have data here

so, cd /usr/ports/graphics/gdal and do 'make missing' let's see what's needed there

we won't need to uninstall pg

since the installed version is 15, and that's the same one we want to build against

ok. try doing make install-missing-packages

that looks like it should be ok to answer Y to

(it won't remove anything)

after that, make missing should come up empty

then you can do a make install

done with message: Message from metis-5.1.0_9: NOTICE The metis port currently does not have a maintainer. As a result, it is more likely to have unresolved issues, not be up-to-date, or even be removed inthe future.

yeah, lots of ports don't have maintainers

ok done

ok, and you're running the make install ?

doing

iirc gdal isn't small, it might take a while

ok

Yup, metis is out of date. The current version is 5.2.1

it's still kind of strange that installing a video editor wants to do GIS mapping graphics, but what do I know

probably some weirdness in opencv plus the fact that the default package builds tend to favour turning options on on the basis that people might need them

are you all guys developers of freebsd?

I'm mostly a user, I contribute to freebsd occasionally but I'm not a committer or anything

How do you contribute?

and why/how you know so much? sys admin?

I maintain a couple of ports, have contributed bugfixes and enhancements to the ports infrastructure, and contributed a few smallish patches to the base system

That reminds me of gnuplot. The default setup brings in wxgtk30, which in turn needs webkit2-gtk3 and gstreamer1.

I happen to know some GIS stuff due to my postgres work, and helping out the PostGIS guys

I see

RhodiumToad there is no money involved in this?

which I why I immediately spotted gdal as the problem here

I don't get paid for freebsd work myself

I see. I wonder how people can invest so much time on opensource. They must be doing pretty well financially from other means already and this opensource engagement would be their passion

eh. never make assumptions about other people's financial situation

gdal still building?

gdal is building; yes

finished