I did it with infocmp

what command exactly

I honestly don't remember

but the weird thing is that it says `alacritty` if I run `toe`

like I need to restart sshd or something

this is nothing to do with sshd

are you using ssh within a terminal session in alacritty?

yes

and you're working with the termcap on the local system or the remote one?

remote one

local is macos

and you're not using screen on either end?

well, I'm not at that point yet, the remote system complains about the terminfo

complains how exactly?

`can't find terminfo database`

if I go to run tmux it just spits that out

what you want to do is to run infotocap alacritty.info >~/.termcap

er no

what you want to do is to run infotocap alacritty.info >>~/.termcap

okay, I'll give that a shot. thanks!

the output of that could use some further work but it'll do as a start

one thing to bear in mind is that removing the linebreaks in a termcap helps with the length limit

weird, it still complains that it isn't found

I actually did take the extra step of adding $TERMCAP to my ~/.zshrc, just in case

don't do that

okay

what value did you put there?

~/.termcap

that's wrong, TERMCAP is set to the entire termcap entry, not to a filename

so unset it to use the file

ah

alright, thanks

On freebsd, is there something similar to Linux overlayfs, where with COW, I can modify/delete a lower-layer file in the upper/merged layer fs? I am aware of nullfs, not sure if it supports it. Thanks!

there is, but it's regarded as a bit buggy

(unionfs)

oh yah, read about it

klu123: will this work out for you? freshports.org/sysutils/fusefs-bindfs

Title: FreshPorts -- sysutils/fusefs-bindfs: FUSE based bindfs implementation way more powerful than nullfs

kenrap: it might. thanks for the pointer!

Good luck!

Hmm, are all sendmail_*_enable="NO" flags required or is sendmail_enable="NO" enough? I think I remember the latter being true for a long while and then at some point it got broken out into individual knobs.

probably I can explan my goal a bit clearer, just in case there's a solution without need to install anything: I've an folder containing subfolders D1, D2, ..D99. I would like to somehow "mask out" folders D1 to D50. So in a mounted parent folder, only D51 ~ D99 are present. Is this doable?

sendmail_enable="NO" is not enough, you can do sendmail_enable="NONE" but I think it's preferred to turn them off individually

Doing this for jails, which have their own logging/metrics/etc services, so don't really need sendmail.

Ah yeah that's right, I used to use NONE instead of NO, but screw it, I'm going to go ahead and individually disable them.

klu123: where's the mountpoint relative to these folders?

Though interestingly enough, /etc/defaults/rc.conf@607 says: "sendmail_enable="NO" # Run the sendmail inbound daemon (YES/NO)."

Kind of confusing, as it doesn't explicitly list NONE as being another potential value.

Hmm, nothing shows up at all under /etc for sendmail_enable="NONE", wonder if that was removed.

see rc.d/sendmail

Oh

it does a case match against [Nn][Oo][Nn][Ee] so it's not very greppable

RhodiumToad: /data. I've root access btw

yup

klu123: I mean, where are the dirs you want to hide relative to the mountpoint

RhodiumToad: like /data/D1~D99. I want to maybe mount them to /data_new, and under /data_new, I only want to see D51~D99

Is there a specific reason that type of match expression is used? I've seen that scattered about in base shell scripts. I mean is it an official style or technical reason or something? My first instinct would be to do something like: `case "${sendmail_enable,,}`; none) ... ;;`

klu123: there may be a way, let me try something

_xor: what do you think that would do?

RhodiumToad: thanks!

Make it shorter, clearer, and most importantly...easier to grep and thus making it easier for me :P

oh

Wait, are ,, and ^^ not posix?

not at all

oh nevermind then

there's no way in freebsd's /bin/sh to do case conversions without an external command

Yeah, ah well.

hence, the idiom of using case and [Nn][Oo][Nn][Ee]

Was just looking up if posix case supports regex and if so would /i work, but apparently not, only supports glob-like patterns. Which is understandable, regex would bloat the implementation I'd imagine.

I sorta got it to kinda work using cap_mkdb to generate a ~/.termcap.db file. sorta.

Does ~/.termcap.db need to be compiled or is it like .login_conf?

it does not need to be compiled

i.e. it should work just as well if you remove .termcap.db, but if both exist it probably uses the .db one

klu123: so there is a way but it's a bit weird? and may require writing a tiny program

klu123: do you expect new directories to be created in the new location, or not?

RhodiumToad: I expect new dirs to appear in the old location, and expect them to show up in the new location. e.g., /data/D101 is created. Then it'll show up as /data_new/D101

ok

so what you can do is this: nullmount /data to /data_new, and then nullmount a tiny filesystem on top of that using -o union (which is not the same thing as unionfs so doesn't have the bugs),

and then create whiteouts on the upper filesystem to hide the dirs that you don't want to be visible

however, for uncertain reasons there seems to be no standard tool to create those

(some scripting languages can do it)

actually you can use another nullmount for the whiteouts, that works too

RhodiumToad: thanks. how do I do whiteouts on the upper layer?

making a whiteout outside of unionfs operations requires root privs, and is done by calling mknod("path",S_IFWHT,0); from C or the equivalent

but the mknod utility doesn't support it

Oh I see. that's why you said no standard tools

scripting languages that have an FFI can do it, I used guile for testing

Got it

Interesting! Yeah I think I can give it a try :)

can I just have one nullmount /data -> /data_new, and do the whiteout there?

no, because the whiteout has to be stored somewhere

and the nullmount has no storage of its own

I can show you my test if you like

yeah that'd be great! thanks RhodiumToad

are you using zfs, btw, or ufs? my test was on ufs

zfs

as far as I know it should work with the lower filesystem being zfs, but I don't know if zfs supports whiteouts

so the upper filesystem might have to be ufs, but it could be a tiny ufs filesystem backed by a file or vdev

dpaste.org/JmDS5 # that's a transcript of what I did; the guile script thing is just how I created whiteouts for D01 and D02

Title: dpaste/JmDS5 (Plain Code)

(unfortunately -o union doesn't have an equivalent of unionfs's -o below, so the new mountpoint dir can't be used to store the whiteouts)

RhodiumToad: great! thanks for the example!

RhodiumToad: also thinking of an alternative using unionfs - does it support whiteout? also if I used a RO nullfs as unionfs lower layer, will unionfs be buggy enough to corrupt my filesystem?

most likely bugs with unionfs would be locking issues

I don't care much about the upper layer being corrupted

I've heard reports of people using it with success, but the warning in the manpage is there for good reason

I think corruption would be unlikely if the lower layer is readonly

yeah man page section is quite scary

ok

whiteouts are fundamental to how unionfs works; it creates a whiteout automatically in the upper layer when you delete something that exists in the lower layer only

I might not use it correctly, but when I do `touch test1/foo; mount -t unionfs test2 test1`. Then `rm test1/foo` I got "rm: foo: Operation not supoprted".

what fbsd version?

oh, and is test2 on zfs? because as I said before, zfs might not support whiteouts

That's a bit embarassed to say lol, but I got 11.4 RELEASE lol

yeah the whole system is on zfs

for testing, you can make a small ufs filesystem using e.g. mdconfig -t swap -s 128MB and then newfs -O2 /dev/mdN where the device is whatever mdconfig said

then mount /dev/mdN somewhere

to keep it around permanently, you'd probably want to create a zvol and use that instead of the md

RhodiumToad: yeah that makes sense! Thanks for the pointer. Let me try...

yeah having a ufs upper layer seems to work

apparently there's an undelete() syscall that can remove a whiteout, but I don't think any tool exists to issue it

oh wait, there's rm -W

sorry, how does that help in my case?

just in case you need to undo the effect of an rm command that hid a lower-layer file

ah, I see! Makes sense :)

OK, lol, just tried rm -W it works :)

quite useful when testing it out, otherwise I'll need to pave the partition somehow

I'm trying to setup an ipv6 router. I have a 2600:* IP for my WAN interface. Should the LAN interface have a global IPv6 address too?

RhodiumToad: btw, is there a way to list all the whited out inodes ?

oh OK, ls -W can display whiteouts

hi

i can't load a fusefs

when I : kldload fuse, i get "kldload: can't load fuse: No such file or directory"

kldload fusefs

ah just found that : fluca1978.github.io/2021/11/03/FreeBSDFuseFSModule.html

Title: Veracrypt, FreeBSD 13 and FuseFS: can't load fuse – Luca Ferrari – Open Source advocate, human being

You need to access lvm ?

Cuz then you also need kldload /boot/kernel/geom_linux_lvm.ko # This will make /dev/linux_vm appear

fixed

nop a borg backup

Cool

repositoy

borg backup repository

Cheers

One more issue after making unionfs with whiteout works: now I want to mount the unionfs to be available inside a jail "X". So in /etc/fstab.X, I added the unionfs mountpoint /data_new to /usr/jails/X/data_new, as nullfs ro. However, when listing /usr/jails/X/data_new with ls, it tries to list all the whiteout items, and complain aobut "no such file or directory"

This has nothing to do with jail, but something to do with mounting a nullfs on top of a unionfs. To repro, first, create a unionfs with whiteouts at /data_new. Then `mount -t nullfs /data_new /data_new2`. Then `ls /data_new2`.

RhodiumToad: wondering if you have any insight on above issue? Thanks!

uh. haven't tried mounting nullfs over unionfs.

RhodiumToad: ok thanks! It seems that I worked it around by stacking another unionfs on top unionfs, instaed of nullfs on top of unionfs.

oooookay

at this point you're way beyond the "beware of dog" sign, so I'm curious whether this ends up working or whether it locks up under load

lol

yeah I seem to have gotten a kernel panic with nullfs lock blah blah

(some humourless person edited that part of the warning out of the manpage at some point, but it's still there in older versions)

I read the flashing warnings in the man page somwhere. I did enjoy it ;)

this is why in my tests I used -o union :-)

it would be really nice if unionfs could be fixed, though.

RhodiumToad: yes, that would be highly desirable!

klu123: oh? What kind of kernel panic?

Hello there

I have an issue with mysql and radius...

while restarting mysql-server takes some time to do its internal maintenance ( idk but it takes ~1-2 minutes to finish )

the radiusd rc contain string : # REQUIRE: NETWORKING SERVERS mysql

but in my guess service listed as started while mysql still not provide a listening socket ...

which makes radiusd service to fail at startup

is there a workaround to this issue ?

right, rc considers REQUIRE to be satisfied as soon as the providing rc script finishes, which usually just means whatever daemon is started, not that it's ready to provide service

RhodiumToad, is there a way to make startup delayed somehow ?

I don't like the idea of polling each minute or so radius service status and restart then

delaying rc scripts by a long time is a very bad idea

I have no other ideas

polling every 10 sec if port / socket is available and then restart service

all my da* disappeared?

ls: /dev/da*: No such file or directory

rsync over ssh too slow. and I can't quickly figure out rsyncd though I did it in past. and arcfour not installed. and -c 0 or no not helping much. so trying to mount sd card w photos on server directly

mac doesn't want to rsync to the locally mounted samba gives a bunch of errors

tried two different card readers

(probe0:umass-sim0:0:0:0): CAM status: Auto-Sense Retrieval Failed (probe0:umass-sim0:0:0:0): Error 5, Unretryable error (da0:umass-sim0:0:0:0): fatal error, failed to attach to device

that's on the newer one. seems bad

usb_msc_auto_quirk: UQ_MSC_NO_TEST_UNIT_READY set for USB mass storage device Norelsys NS1081 (0x2537:0x1081)

nerozero: I haven't used MySQL/MariaDB in a long time (aside from services that require them), but I think I remember there being an event hook that is triggered when the server is ready to respond to queries. I'd imagine you could use that to start radiusd (or restart it).

or send SIGHUP, etc.

da0: <Kingston FCR-HS219/1 9722> Removable Direct Access SCSI device is the other one, older usb2 reader. doesn't want to mount

_xor, thanks

qmr: do they show up in usbconfig?

qmr, rsync slow ???

nerozero: yes. like 12-13mbyte/s. the wifi is much faster.

qmr, I have a gigabit network and somehow it can utilize its fully

RhodiumToad: stil have the newer one plugged in yes it's there. tha'ts cute it shows current for each port nifty

its highly dependent to a file size

large files - works faster

nerozero: mostly video files

What file system is the card formatted with?

you can also use sshfs

fat32

are you sure it's fat32 and not exfat?

would be nice if mac and samba would get along. my next try was going to be nfs maybe

I think so? camera splits at 4gb. but I can check

sshfs / scp

and how did you try to mount it?

not sure why that would be faster than rsync over ssh but I could try. I prefer rsync to know for sure transfer was complete too

12mb is OK speed for a usb thumb drive

not on usb3 it's not

qmr it would be the same speed

What error are you getting with rsync? I've found that flags like --no-owner --no-group --noperms helps when there's a feature mis-match between the src and dst file systems (that's assuming you don't need those to be set though).

use a gstat to see which drive is fully used

in case of high fragmentation and low space ZFS can become very slow

RhodiumToad: mount with a bunch of args. got not ready or read only

low free space

and test destination drive is not a SMR drive

RhodiumToad: I was reading recently that, practically speaking, USB2 is going to be way more reliable than USB3 due to real-world implementation differences between hardware (couldn't follow the full article/comments as it was beyond my USB-specific knowledge). Not saying it applies in this case here being discussed, but that sticks in the back of

my head now as something to consider. Not sure how true it is though.

qmr: if the device is read-only, did you include the -o ro or -o autoro option?

_xor: something about extended permissions or something? rsync: failed to set permissions on "/Volumes/media/incoming/Downloads/20190817_130032.jpg": Invalid argument (22)

RhodiumToad: -o ro yes

the usb card readers I have present two separate da* devices, one for each slot (sd and microsd)

qmr: If you don't need or care about the permissions, then try --no-perms --no-owner --no-group.

qmr this is on in case of fat

qmr: can you show the exact mount command and the exact error please

and also consider that in heavy use Flash memory become unusable and slow

it's not an issue with the card, it transfers quickly fine

RhodiumToad: shouldn't there be /dev/da something?

(da0:umass-sim0:0:0:1): fatal error, failed to attach to device

qmr, this is obviously a hardware issue

somehow USB cannot be properly attached

try resetting the device with usbconfig

went back to the usb2 reader this time I put the reader in for a moment, then put the card in some moments after. it shows a bunch of da*, one of which has a 119G storage

[root@frisbee /media]# mount -t msdosfs /dev/da1s1 sd/ | mount_msdosfs: /dev/da1s1: Invalid argument

fstyp /dev/da1s1

ooh it's exfat

:)

dunno why camera splits files then

fusefs-exfat-1.4.0 <- ? -t exfat doesn't work

yes, you need that package

(and use mount.exfat-fuse)

oh I did mount.exfat

same thing, just a symlink

waiting for this to hit a video file to see what kind of throughput I'm getting. I have a small usb3 card from a dumpster PC to put in server I could use the usb3 reader then maybe

30-35mbyte/s. but getting a lot of scary warning messages

what messages?

seems it's choking on any video file bsd.to/v3NT

Title: dpaste/v3NT (Plain Text)

omg, host doesn't support a port argument?

hi

I am new to freebsd. Which iso is best for laptop

whioch USB3 are you using, try the one located directly on motherboard

I meant is there any stable version of beta version exist?

node1, 13.2. 12.4 won't be supported past the end of 2023.

So which is good for me? i am planning to switch from linux to freebsd.

qmr, I recall I had some issues with USB3 ~ a year ago, USB3 doesn't recognized any devices while some other usb ports has been plugged in ...

it was on a Asus motherboard

just in case

node1, what CPU does your laptop have?

It's AMD

Then I'd say download.freebsd.org/ftp/releases/I…FreeBSD-13.2-RELEASE-amd64-dvd1.iso.

I'd use the memstick or mini-memstick image myself

Write to memstick, Change volume mount option to RW, and now you have full functional live system :)

for small values of "functional"

ok ok :)

Quite useful when system fails, can contain OpenVPN which allows to connect directly to remote server automatically

then it is possible to ask someone to just boot from a stick - and suddenly you have full access to a remote machine

A dead machine

RhodiumToad, so would I, if it were me doing the install.

Given a command pipeline, what would be the "correct" way to format stdin for one of those commands and pass it on to stdout? I'm looking at possible ways to do it with print, sed, awk, etc.

Basically, I'm doing a DNS lookup and the command outputs JSON, but it's loosely valid. It's not wrapped in an array, just newline delimited objects. I need to wrap the whole thing into an array.

So you'd need to start the array, maybe add a separator like ',' after each element, and end the array?

I'd use awk if so. BEGIN {...} and END {...} for the array start and end, and a patternless action for the added element delimiter if any.

Or perl, if that's in your "etc".

_xor, python dns lookup and then serialize to json ?

or maybe even php if already installed

in python: socket.getaddrinfo

or socket.gethostbyname / socket.gethostbyname_ex

_xor: i think libucl, and by extension uclcmd, might be able to deal with that

uclcmd is like jq for ucl. ucl can read things that are json, but also things that are *like* json (which UCL is.

hrm. so "ifconfig wlan0 list scan" says that my AP has a -79 signal, and -96 noise... which doesn't sound great. but wifianalyzer on my android phone right next to my laptop says -60 dbm

Hi! I managed to pass my GPU card to a bhyve machine. I installed the drivers and have video output. However, I also need a mouse and keyboard. Can Bhyve pass a single USB device to the virtual machine? I cannot pass my entire controller because when I do this, the machine freezes.

Ronis_BR, I found it easier to get a dedicated PCI USB card for passthrough.

davisr: hum, thanks! Unfortunately this is not an option for me. The server is under warranty period and we cannot open it :(

Ronis_BR, oof. Maybe just VNC into the virtualized guest?

davisr: but if I VNC, how can I see the video output by the GPU?

oh, maybe RDP!

Ronis_BR, I guess that would depend on your guest OS. Usually, one can configure a VNC server to attach to an existing display. But since you said RDP, I'm guessing your guest is Windows?

davisr: yes, it is!

Ronis_BR, I'm not familiar with Windows (yuck) but I'm pretty sure it wouldn't render a separate desktop. You could just RDP or VNC into it, and it would render the screen via GPU and mirror its output.

davisr: the problem is that now I have two video cards. I need to disable the one created by bhyve

Ronis_BR, but I would recommend against wasting your time with passing through a specific USB device.

davisr: yes, I see this will be bad. I really prefer to use remote desktop if I managed to do this

davisr: do you know if there is plans to support individual USB pass to the VM?

davisr: will this work:

Title: Placa Usb 3.0 Pci Express 1x Pc Computador 4 Portas Saidas | MercadoLivre

Ronis_BR, I don't think bhyve is designed to do individual USB passthru, only PCI. I will advise that not all USB cards work! I had to try a few different ones. Mine uses the ASM1142 chipset.

thanks!

Someone in #alacritty told me my terminfo problem is an issue with ncurses dependency (base versus port) and I'm sad about it because he was right.

installed zsh/tmux/ncurses from pkg and nothing really behaved as expected. installed zsh/tmux/ncurses from ports and everything works great.

ncurses or terminfo data?

nero: no amount of invoking `tic` would convince my terminal that `alacritty` was a valid configuration until I installed everything from the ports tree.

davisr: the RDP is working perfectly! I just needed to change graphics option to NO in vm-bhyve after the installation.

davisr: thanks for the help!

so i'm familiar with file reporting tools for deduplication like fdupes, or just manual md5 comparisons. anyone know a tool to compare two directory trees and identify the duplicates/differences?

i really think several whole directory trees are duplicates, so telling me 10k files in each are dupes doesn't help, i need to summarize by parent dir

Do duplicates in your use case have the same filenames (even if different parent directories)? Or do you also need to spot that foo/bar and baz/quux are duplicates?

V_PauAmma_V: frequently yes

Frequently the latter, you mean?

most of the filenames are similar, yes

i have 3-4 copies that diverged of a directory tree of a few gb and 10k files each

trying to merge

I'd start by postprocessing "diff -Nqrs", then.

"You can not have the same branch checked out in multiple worktrees." - gah, is that still true about git?

Hard luck doing a simple thing. There is no "working way". How exactly do I change mouse pointer to be bigger in size in freebsd + awesomwm?

so, the reason for shit rsync over ssh was actually because I went to wpa1 + 2 and I guess that disables channel steering? so I was connecting at 2.4ghz instead of 5ghz. I think I will switch to 2 ssids

~40mbyte/s now which is ... acceptable I guess

discovered something else funny though iperf to local server is worse than internet speed test by a lot. going to need to look into that

be nice if I could get that card reader to work properly and do an auto ingest setup. plug card reader in then get an alert when ingest is complete then sort the files

regading the usb3 card reader saw some issues about usb "quirks" on bugzila with that model