My shell script wont work on freebsd :( i guess it has too much bashism.. (read -a M<<<"$(echo 'scale=62;obase=36;2469958312622/2821109907455'|BC_LINE_LENGTH=0 bc -l|cut -c 2-)";L=${#M[@]};for i in ${M[@]:0:$L};do printf %x "$(($i+55))";done)|xxd -r -p;echo

(it uses math to prove VIOLENCE really is the answer sometimes)

rustyaxe: does it do 'rm -rf /' when decoded?

No

it shows VIOLENCE several times

i suppose i could do one that does that but it'd require another subshell and significant maths

You can try it on wolfram alpha, wolframalpha.com/input?i=2469958312622%2F2821109907455+to+base+36

Title: 2469958312622/2821109907455 to base 36 - Wolfram|Alpha

click more digits some times, it repeats forever.. math is fun :)

the <<< is a bashism

xxd I think is a linuxism, it's definitely not a standard utility

the ${#M[@]} is a bashism (or kshism)

the rest looks like standard shell

hmm it seems xxd isn't so common :( that's troublesome

what's it do?

its somehow related to vim; in this case it converts the result from hex to .. something. It's been a minute since i wrote that

the way bc formats its output is... broken, so i used that to fix it

looks like xxd converts to or from a hex dump

56494f4c454e434556494f4c454e434556494f4c454e434556494f4c454e434556494f4c454e4345

Ya from hex

If one makes a script with enough CPU threads, they could find more of those. I just remember finding that one years ago when wolframalpha was young

Learn what an Alabama Hot Pocket is by reading this spellbinding story of the steamy hot meeting between Mexico City Mayor Claudia Sheinbaum and Campeche Gobernor Layda Sansores. Edify yourself on Mexican Politics and Alabama Hot Pockets today! justpaste.it/Sheinbaum_Alabama_Hot_Pocket

Title: Claudia Sheinbaum, Layda Sansores, and the Alabama Hot Pockets - JustPaste.it

I was really bored and figured out that 2469958312622/2821109907455 when taken out enough places (it repeats infinitely) and turned to base 36 gives the desired result

This seems like spams

I'll poke at it some more and avoid the <<< and xxd.

is there an equal to ${#[@]} in sh?

nope. there are no arrays in standard shell

however, you can usually use the positional parameters as a workaround

i mostly did it this way to obscure a bit what is going on, of course. So even one even a little familiar can see its using bc to do a calculation... and.. things with it lol

consider:

set -- $(echo ...);for c do printf ...;done

set -- $(echo 'scale=62;obase=36;2469958312622/2821109907455'|BC_LINE_LENGTH=0 bc -l|cut -c 2-);for c do printf %x "$(($c+55))"; done

56494f4c454e434556494f4c454e434556494f4c454e-sh: arithmetic expression: expecting EOF: "1\+55"

Close! im sure i can figure out a way to avoid xxd there, its just hex to ascii after all

BC_LINE_LENGTH isn't being respected (it's not a standard option)

looks like newer fbsd releases have a bc that supports that

hm

on my system the manpage says BC_LINE_LENGTH works but in fact it doesn't

hmm there's -L but thats not portable

*tries it on bash on linux vm as changed*

time to dig into the source

still works on bash on linux, closer to working on sh on freebsd. Maybe for another evening ;) its a fun distraction, especially if i can find more words

i found the #s wrote in my notebook from 20+ years ago... and this too -- -72.01,168.57778 in google earth.. zoom out a bit :)

wtf!

somewhere, 1 is being subtracted from line_len

anyway, I found the bug

hmm?

the bug in bc that makes BC_LINE_LENGTH=0 not work

at least as of fbsd 13-stable

looks like it's already fixed upstream

not fixed yet in -current as of the last time I fetched it

Ah, its mostly a toy anyways, just wanted to improve it to be a bit more portable. Thanks for the tips, it still works in bash, i'll poke at replacing xxd with something portable another evening ;)

I have a weird thing where some fonts aren't rendered correctly in firefox

hoping somebody has seen this before and knows why

only a few fonts and sites, it may have been broken for weeks even

are you settings fonts with fontconfig

I've tried disabling h/w accel in firefox, and similar settings

morpho: probably not, I'm a tmux + defaults kinda person

nvidia card

as far as I know, firefox uses fontconfig, its default

that is a very strange error

is this wayland or xorg?

There could be a mismatch between the text encoding & lack of suitable font; or, incorrect encoding might been specified. dch, run "firefox >| log.firefox 2>&1" to examine if firefox spits out something. There may be some font-debug option IIRC

xorg

parv: good idea, I'll try that this arvo

in .config/fontconfig you can declare a 'default' font which will be the fallback for firefox. Noto Sana Mono has pretty good coverage

the best I know of

is this firefox or firefox-esr?

dch: can you link to any of those sites?

yuripv: no, this one is an embedded term inside oracle cloud.

it needs a login and stuff to get to it

but if I find another one I will ping you, thanks

Title: Element

also has this problem, but only for the favicon in the browser tab row

hmm maybe there's some font cache or something that does this

I'll see what I can nuke after restarting

what is the preferred way to run google chrome on freebsd (required for work stuff)? i was thinking either a linux VM with bhyve or a jail. i don't want it funking up my main system

dch: to recreate the cache, try `fc-cache -fv` (-v shows the paths as well)

trev: you cannot use chromium?

Lovis_IX: no, i need to login for the sync (work controlled browser). i am leaning towards VM

don't think it's possible with jail

trev: ok. So bvyve or virtualbox should be some solutions.

trev: its possible with a linuxlator jail

dch: what do you mean with "linuxator jail"? A bvyve jail?

trev: see github.com/mrclksr/linux-browser-installer and maybe github.com/hnhx/freebsd-steam has some help

Title: GitHub - mrclksr/linux-browser-installer: Script to install Linux browsers under a Linux chroot on FreeBSD

freebsd has a partially complete linux ABI called "linuxlator"

it can run many linux programs OOTB, with some caveats

not all syscalls are emulated or provided

but you can play steam games, and browsers, via this

anyway trev I'd just use www/ungoogled-chromium from ports directly, and I'm sure you will be fine

yuripv: thanks for the fc tip!

or just www/chromium, it will do the sync fine, IIRC, it used to in the past.

failing that, linuxlator and that will also work

jailing X apps is possible, vermaden.wordpress.com/2021/12/15/secure-containerized-browser honeyguide.eu/posts/pot-throwaway-firefox both good approaches

Title: Secure Containerized Browser | 𝚟𝚎𝚛𝚖𝚊𝚍𝚎𝚗

thanks friends

i think they removed the sync from chromium. work needs to "have total control" over the browser -- whatever that means

i guess they can nuke it or monitor it, no idea

thanks dch.

trev: good luck

Is FreeBSD 14 stable enough to be my daily driver? Asking because I'm doing literally all my work from this machine but my new gpu is only supported in 14.

phryk: what gpu?

rx 6600xt

rx6600 looked like an HP INTEGRITY at first sight.

phryk: so long as you don't do anything funky with ZFS, you should be good…

i was hoping for the kernel changes needed for newer drm to be backported to 13, but it doesn't look like it.

meena: eh, I don't have zfs on this machine and am deprecating it on the one machine i still have it.

phryk: why?

(as in: I'm curious, not: HOW DARE YOU?!)

too much pain for zero noticable advantage.

what will you use instead? UFS?

what else is there? :O

way too many traps, completely different way of managing it compared to literally every other filesystem and i don't think there's really any feature i want that i can't have with ufs… encryption, mirroring, snapshots, etc. are all possible with ufs and you don't have to learn an obtuse new way to manage things…

i mean i used zfs for over a decade and i still stumble onto stupid mounting issues that just shouldn't exist.

phryk: you should be fine. also, you are welcome to bug me and ask for help anytime

each to their own with zfs, I've had zero problems with it, and run it on everything from laptops to servers. but I use it at work, not as a hobby, so I had time to get familiar with it

dch: that's nice but i kinda assume that if i run into problems, those will mean i won't be able to ask you.^^

lol

there is that

you would have to hunt me down at home

throw bricks at my window

i have a rough inkling where that's is. like 1-2h walk.^^

slower if you bring many bricks

true.

dch: i've had a good bunch of performance problems with zfs. worst was when i accidentally activated dedup on a machine with 6g ram – slowed down r/w rates to like 200k/s. fixing that was a major hassle (essentially rewrite everything, wouldn't even have worked if i didn't have proper mirroring set up). for another issue, i had help from people here about a dozen times over ~2 years and it just

remained a complete mystery. never had any such issues with ufs.

dedupe is the enemy. the #1 feature that should be removed IMHO from zfs

and yeah that is brutal to fix

I have also added an extra 10TiB as a *stripe* instead of a mirror to an existing drive

that was annoying to fix

checkpoints people

they can rescue all these fuckups and more

yeah, but it's not the only weird thing to happen. i still have issues with a couple zfs datasets not being automatically mounted and the like…

they are the best

tykling: you mean snapshots?

no, snapshots are of data, checkpoints are of pool config

so when you fuck up and enable dedup or add a 10tb stripe you can undo it

yeah, never heard of those. were they a thing in 2011 already?^^

zfs is not to blame for admin mistakes, but at least it can help fix them :)

i mean, that's sorta true. but no other fs has all these weird traps. IMO the cli should at the very least warn and do double opt-in for that kinda stuff.

you've had since 2007 to put zfs on top of a truncate(1)'d file, and learn how to administrate it.

ufs and every other filesystem has just as many gotchas, they're just different gotchas.

phryk: its not obvious initially but almost all z* commands have a `-n` dry-run option first

only difference is, you're familiar with them.

and a `-f (fuckit)` option for when you want to YOLO your data

it's the same stupid argument as "i like pf because i used that first"

debdrup: i literally have no idea about ufs gotchas and apparently never ran into them.

debdrup: pf has an objectively better config format than $OTHER firewalls, and an active maintainer ... there is that

i used zfs on my first or second freebsd install in 2011, so i really don't think the argument that i only like ufs more because i used it first has any validity.

dch: it's got a better config format for you, since you used it first

i'm way more familiar with ipfw, since i used that first - so to me, ipfw is better

debdrup: but ipfw requires numbered rules. that's like 1985 config style

numbered rules are an advantage

anyway your point is valid, we know what we know, and thats confirmation bias

FORTRAN CALLED AND WANTS ITS CONFIG FILE BACK

etc…

until you want to insert a rule between 999 and 1000

theres some cool stuff like dummynet that IIRC is only available with ipfw

the feature set is different

dch: you've got 2^16 rules - spread them out a bit more.

i only ever used pf and i think it's… well not awful per se, but definitely not great.^^

as a refuge from iptables on linux, I heartily disagree.

there are lots of things I'd like pf to do better

my #1 issue with UFS is no on-disk/in-block checksums

and that will never be fixed.

dch: that reminds me, have you seen that pf bug: bugs.freebsd.org/bugzilla/show_bug.cgi?id=268717 ?

Title: 268717 – [pf] [ipnat] rdr rules don't work for traffic originating at localhost

and imho, SU+J are not as "rugged" as zfs transaction groups

I switched from MacOS to FreeBSD because of bitrot that went unnoticed for several years, and backups got cycled out, and I lost a bunch of very early photos

phryk: silent data corruption, lost+found folders, the need to fsck things occationally, static inodes, no snapshot when doing soft updates + journaling, no ability to detect bugs in device drivers (while they're unlikely, they're not impossible), can't deal with phantom reads/writes, can't handle dma parity errors, and don't even konw about misdirected reads and writes, and probably a whole host of

other things

those are issues with traditional filesystems like ufs

debdrup: on the zfs side you have a filesystem that's extremely sensitive to fragmentation, introduces it a lot and doesn't offer any tool to defragment. that killed my performance more than once.

anyway phryk you should totally try current. if you were using zfs then you could use boot environments and switch easily between current & release as you like. This is not so easy with ufs, but its very straightforward with zfs.

pros and cons

my added-stripe-not-mirror admin fail was before zpool checkpoint existed sadly

the root pool on my homeserver has… 60% fragmentation. and zfs people keep telling me that it's a negligible issue @_@

time for me to pay attention, /etc/inetd.confg is not a thing

phryk: then there's things like fragmented storage (meaning you have to split stuff across different disks - unless you're using graid, which has its own problems if you're doing raid5-equivalent stuff because of write holes), and that there's simply too many limits on the expandability

phryk: fragmentation is absolutely a problem over time. depending on pool layout it can be easy to fix

debdrup: i use gmirror.

phryk: zfs used to have a problem with free space fragmentation (which isn't the kind of fragmentation you're talking about), but it's much less of an issue now than it used to be, and if you free up space it stops being a problem.

(zfs send stuff | zfs recv stuff) will clean it up, but obv you need to unmount/remount

phryk: good for you, but that doesn't address the issues with raid5 write holes; the equivalent level in raidz doesn't have write holes.

but this whole "ufs has no problems" is just utter nonsense.

then use zfs. i'm not here to evangelise you, lol.

imma get me some food

they're designed to address wholly different things; ufs was designed in the very early 80s to have somewhere to put data on. zfs was designed because storage subsystems _lie_

I thought the whole 'last word in filesystems' thing was the main drive.

i.e. coupling what we think of as a block device and filesystem into one, glorious, whole

i never backup anything. what is the right FS for me?

can't lose data if you don't have it

👍

i did recently watch a ZFS noobie video from EuroBSD. seemed kind of neat, but i would never use any of the features unless i have to replace a drive

Robert C's comment in bugs.freebsd.org/bugzilla/show_bug.cgi?id=263377 -- why notify wireless@ as "igb" driver looks to be for RJ45-kind?

Title: 263377 – [NEW PORT] net/intel-igb-kmod: out-of-tree driver for igb(4) from Intel-provided source

can somebody help me delete this messed up route: gist.github.com/340135858b657931b2b4db878222ab21

Title: gist:340135858b657931b2b4db878222ab21 · GitHub

this route was added with `route add -net 192.168.11.0 -interface lo0 192.168.11.1` but the result is rather surprising.

no, wait

`route add -net 192.168.11.0/24 -interface lo0 192.168.11.1`

route -v del -net 192.168.11.0 -netmask 192.168.11.1

i need to test this on CURRENT

the "mask" reported by route makes no sense

okay, i can add the route in parts: `route -v add 192.168.11.0/24 -iface lo0` followed by `route -v change 192.168.11.0/24 192.168.11.1` but trying to combine that completely shits the bed

i am utterly confused

people who want to become maintainers for iocage, please apply: iocage/iocage #1289#issuecomment-1532945748

Title: Last Accepted Pull Reqest was Oct 2021 · Issue #1289 · iocage/iocage · GitHub

1289 – errno breaks in thread-safe c++ compiles bugs.freebsd.org/bugzilla/show_bug.cgi?id=1289

dubiousness: ZFS combines logical volume management and filesystems into one, and what makes it the last word in filesystems is that it was written with the assumption that all hardware-based storage subsystems lie, which up until that point wasn't something anyone had considered (and I'm not sure anyone's taken it as seriously as Ahrens and Bonwick did).

I don't know, I have it on ignore.

I don't, and it does, and that can be confusing (briefly for me, unsure whether or how long for others).

Is there any benefit in running ZFS with a single drive?

Or do you need multiple drives to really see its magic?

RoyalYork_: there's a few downsides to running it on a single drive, and no real benefit.

Peace of mind and future-proofing for me. If I ever need more than the storage I have on my laptop, I'll sacrifice the CD/DVD, replace it with another disk, and stripe.

You need to enable ditto-blocks pool-wide to be able to recover from UREs on the data (metadata has ditto blocks by default), and that in turn means that every write takes double the amount of time (because it has to write it twice).

Most people forego that, because you still get informed about the actual data loss of an URE even if you don't have ditto blocks, and it's simply not necessary to have two disks or ditto blocks if you have some other storage that you back up to frequenly enough.

can someone review if this makes sense: canonical/cloud-init #2165/commits/6640f00f0e253da1351e251328c4dde4aececa01 ?

Title: Ephemeral Networking for FreeBSD by igalic · Pull Request #2165 · canonical/cloud-init · GitHub

I had to use two steps for what Linux can do in one step. plus, we can't do source selection.

If I'm reading the python correctly, wouldn't line 25 make more sense this way: if (( gateway ) and (gateway != "0.0.0.0")): ? just so the meaning is more clearer.

why would you add those ()? well, why would you do if gateway and gateway != "0.0.0.0": instead of just if gateway != "0.0.0.0": ?

that's the same style as elsewhere, so that's not the question

oh, wait, you need to test that gateway is thruthy, I was stupid there

So, from what i gather, the main data this is used with is going to be: static routes from DHCP

hrm…

I love messing with people's patches: bugs.freebsd.org/bugzilla/show_bug.cgi?id=271243

Title: 271243 – sysutils/tuptime: Fix cron file

there I also don't see a point in using ()s (in the first version yes, but why in the simplified version?)

nimaje: good point

RoyalYork_: I run ZFS on my single disk laptops (FBSD & Ubuntu) because block based backups are dead simple and fast with zfs snapshot replication

not to mention boot environments

jgrafton, thanks for the insight.

I had installed zfs originally without understanding what it actually was

RoyalYork_: take a little time to explore it. I really love ZFS, even with a simple disk, like jgrafton

zfs good status: checks out.

I noticed a small documentation bug, and I guess since I'm already typing this, I'll mention it here before I put it into bugzilla. On freebsd.org/releases/13.2R/announce there is the following text: "Please note that 13.1 will be supported until three months from the 13.2 release date, which is yet to be scheduled at the time of this writing."

Title: FreeBSD 13.2-RELEASE Announcement | The FreeBSD Project

While 13.2 has definitely been released. :B

IRC makes for a terrible bug tracker.

true...

( Man "proj-data" is huge ... )

( ... 676 MB of whatever packgae that is )

parv: it's a GIS thing

it gets pulled in as a test dependency of some otherwise trivial package, iirc

(if you're finding yourself downloading it and you're not doing any GIS stuff, that may be why)

RhodiumToad, Thanks. Also possible I might have added when was imagining myself trying GIS (& haven't gotten around that); in any case, "pkg info -r" does not list dependent packages.

Ok, removed "proj-data"(for now)

hm, it used to be a dependency of graphics/proj I think but no longer is (there's a commented out line)

I had "proj" installed at one point, as for reasons alluded to earlier. Yeah, that could be it. Thanks RhodiumToad

I sometimes do pkg autoremove to get rid of old stuff