<cole-h>
Oh, I used `date +%G` instead of `date +%Y` lol derp.
<abathur>
hehe
<supersandro2000>
zeroconf tests are fun: first they fail with `OSError: [Errno 40] Message too long`
<supersandro2000>
and the it dumps the entire debug log to my terminal
<supersandro2000>
in hex escaped
<supersandro2000>
sanic has a very nice progress bar and emojis and there debug log spammed the terminal for a good 30 seconds....
endformationage has quit [Quit: WeeChat 2.9]
hexa- has quit [Quit: WeeChat 2.9]
hexa- has joined #nixos-chat
<abathur>
oof
metalune[m] has left #nixos-chat ["User left"]
waleee-cl has quit [Quit: Connection closed for inactivity]
cole-h has quit [Ping timeout: 256 seconds]
fnlaai has joined #nixos-chat
<fnlaai>
hi
<fnlaai>
how r u guys?
fnlaai has quit [Ping timeout: 245 seconds]
<bbigras>
hi 👋 doing good and you?
<lovesegfault>
,locate avconv
<{^_^}>
Found in packages: libav, libav_12, libav.bin, libav_0_8, libav_12.bin, libav_0_8.bin
<lovesegfault>
hmm, can any of you run `gource`?
<lovesegfault>
I get gource: SDL initialization failed - GLEW Error: Unknown error
<bbigras>
gource: directory not supported
<lovesegfault>
gotta do it in a nix dir
<lovesegfault>
s/nix/git/
<bbigras>
it's working
<lovesegfault>
hmm
<lovesegfault>
I wonder why it's broken here
<lovesegfault>
Ah, I had to unset SDL_VIDEODRIVER
<supersandro2000>
make a wrapper?
<supersandro2000>
it took me like 20 minutes to find all packages which try to build ceph...
<supersandro2000>
but hopefully I got all and I can now skip building ceph which takes forever
<eyJhbV2>
,ping
<{^_^}>
pong
<eyJhbV2>
Damn it, I hate when you get "banned" from channels, if your VPS is somewhat unstable, and then you are unable to switch back to your nick
<samueldr>
yeah, channels forbidding nick changes, mixed with channels forbidding unregistered users to join causes a bunch of issues
eyJhbV2 is now known as eyJhb
eyJhb has joined #nixos-chat
eyJhb has quit [Changing host]
<samueldr>
not only with unstable VPSes, but netsplits too
<eyJhb>
One should be able to ban those channels.
<eyJhb>
Anyways, back to exam stuff...
spudly1 has joined #nixos-chat
spudly1 has quit [Read error: Connection reset by peer]
spudly1 has joined #nixos-chat
spudly- has joined #nixos-chat
spudly has quit [Ping timeout: 272 seconds]
spudly- is now known as spudly
<samueldr>
this intel-based tablet is cursed
<samueldr>
I tell you, cursed!
<samueldr>
the bootloader for android, which is an EFI program, will change the ACPI tables wildly
<samueldr>
the *touchscreen* doesn't respond the same way, even!
<samueldr>
totally different calibration data, or something along the lines
<samueldr>
a full xhci usb device is not shown to non-android OSes
<samueldr>
this cherrytrail tablet somehow just hides the OTG XHCI controller from non-android systems
<samueldr>
probably not to confuse windows
<samueldr>
I've had this tablet for years, and just now I've amassed enough knowledge to dig further, find a kernel mailing list thread for a similar situation, where the user observed the same cursedness
<samueldr>
(disregard about the touchscreen mapping, user error)
<samueldr>
still, when launched directly from the UEFI firmware, no way to use the usb gadget... launch from the android bootloader and it works
<samueldr>
good thing I made a whole thing to produce android boot images recently
lunc has joined #nixos-chat
tilpner has quit [Remote host closed the connection]
tilpner has joined #nixos-chat
FRidh has joined #nixos-chat
<Ashy>
samueldr: slightly unrelated, have you tried edk2 on the rockpro64 yet?
<samueldr>
I don't have a rockpro64
<Ashy>
ah ok, my memory must be failing me then
<viric>
samueldr: I had an intel-cpu tablet and the bios was nightmare.
<viric>
And the linux drivers part wasn't good either. I requested the GPL pieces of the android part and I couldn't get them.
<viric>
The "manufacturer" said that they couldn't distribute the GPL because of Intel conditions
<viric>
and told me to discuss with Intel about that. But Intel would say I should ask the manufacturer.
<samueldr>
sounds about "right" with that kind of manufacturers
<samueldr>
and don't worry, the bios is a nightmare on that one too
<viric>
And from some discussions in #intel-gfx on irc I understood that Intel could yes be doing that.
<samueldr>
and the manufacturer never released the sources to android either
<samueldr>
uh?
<viric>
The manufacturer said he didn't have the code, because Intel denied it to him
<samueldr>
odd
<samueldr>
but yeah, the kernelflinger android bootloader messing with the DSDT was a new development for me
<samueldr>
I tried comparing the DSDTs but it's a bit too thick for me
<viric>
I expected the Intel to be a platform that would stand time much better (for linux common support) compared to ARM boards
<samueldr>
haha, the main reason it's "that good" is the install base
<viric>
but that was not very correct, to my surprise.
<samueldr>
and wintel sure did help
<viric>
Moreover it run android+windows10, so I was sure Linux would run there.
<viric>
It run, but very poorly.
<samueldr>
I kinda knew when I got it that it wouldn't be that easy
<samueldr>
(I did some research beforehand)
<samueldr>
but I also saw that there was some fast progress at the time on what was needed for my uses
<samueldr>
and now that I've resolved the mystery of the missing XHCI controller, all things are fine
<viric>
at the end it broke and threw it.
<samueldr>
did it just stop booting?
<viric>
no, broken to pieces
<samueldr>
ah
<samueldr>
the bios for those is **odd**
<samueldr>
well, for the one I have
<samueldr>
and sometimes just decides to tie itself into a knot
<samueldr>
and not boot
<samueldr>
unless you "reset" it by holding power for a good minute and a half
<samueldr>
which resets the bios settings fresh
<viric>
It seems that a tablet or mobile phone cannot be a long-time supported platform even for linux world
<viric>
I can't understand that much; even sound blaster is still supported.
<samueldr>
I don't think that's a foregone conclusion
<samueldr>
the problem is those don't really work *with* Linux, but *use* Linux
<samueldr>
without trying to contribute back
<viric>
but is that code so bad that it cannot get into mainline?
<samueldr>
*and* they don't even try
<samueldr>
a specific contributor (jwrdegoede) has seemingly taken a liking to baytrail and cherrytrail platform, and is writing fresh code (not like there's downstream code) for them
<viric>
baytrail was the platform I had, I think
<samueldr>
most likely
<samueldr>
there really was only baytrail and cherrytrail
<samueldr>
cherrytrail is mostly baytrail++
<samueldr>
the xhci controller name advertises as baytrail
<viric>
so much effort lost
<samueldr>
yeah :/
<viric>
Did someone get the GPL parts for all baytrail pieces?
<samueldr>
no idea
<viric>
that's what I failed to get
<samueldr>
I don't know the details, only that mainline works well enough
<samueldr>
(on cherrytrail, but probably about equivalent on baytrail)
<viric>
intel-gfx didn't work, speedstep didn't work iirc
spudly has joined #nixos-chat
spudly has quit [Changing host]
<viric>
well I hit that about 4 or 5 years ago
<samueldr>
yeah
<samueldr>
sounds about right for that time frame
<samueldr>
IIRC when I got it I *had* to use the fork from jwrdegoede which added a lot of stuff
<samueldr>
(all into mainline now)
<viric>
so the trick is to get things into mainline - then they are rarely taken otu
<viric>
out
<samueldr>
basically
<samueldr>
but it's also rare that it goes into mainline for only one bit of hardware
<samueldr>
you can observe what the chromeos project does for a success story with mainline (mostly)
<samueldr>
all their work *has* to be contributed back to mainline as a goal
<samueldr>
meanwhile you can look at android OEMs in the corner eating paste :(
<viric>
I see.
<viric>
Is the raspberry basically all mainline?
<samueldr>
I see you're not in #nixos-aarch64 ;)
<viric>
I don't have any aarch64
<viric>
all 32-bit
<samueldr>
yeah, joking a bit
<samueldr>
(but we do also discuss 32 bit ARM)
<samueldr>
I ranted a bit about how raspberry pi is just as bad a vendor as other vendors
<viric>
ah. well, it's up to broadcom I guess
<samueldr>
maybe worse because they create vendor lock-in with their boot flow, and incompatible kernel ecosystem
<samueldr>
not sure it's up to broadcom really
<Ashy>
edk2 is actually pretty nice on the rpi4
<samueldr>
I assumed they do a lot of work in-house for kernel development
<Ashy>
it's weird and nice having a uefi menu on an sbc
<samueldr>
Ashy: yeah, EDK2 on raspberry pi 3 also is really good
<viric>
what is edk2?
<samueldr>
"the" reference UEFI implementation
<samueldr>
and the best UEFI implementation probably
<samueldr>
well, no
<samueldr>
tianocore would be
<samueldr>
EDK2 is the SDK IIRC
<viric>
what is the plot you see in raspberry then? just a bad vendor like others?
<samueldr>
vendor, with possibly not "master mind villain" ambitions, probably more about not caring to work with mainline better
<viric>
I was looking for a device that would allow me to run pulseaudio (for echo cancellation) and h264 encoding+decoding for 720p webcam.
<viric>
with hdmi output.
<viric>
many devices could do that, but all of them have software disasters.
kalbasit has quit [Ping timeout: 240 seconds]
kalbasit has joined #nixos-chat
__monty__ has joined #nixos-chat
ottidmes has joined #nixos-chat
<viric>
I have many external HDs with luks. And I do too much handwork to mount/umount them. What is the proper schema?
<viric>
can luks partitions be named?
<viric>
maybe I should be using gpt for that
<patagonicus>
There's /dev/disk/by-label, which uses filesystem labels given to mkfs, and /dev/disk/by-uuid (unique identifiers for each fs), and /dev/disk/by-id, which should use "dm-name-${luksName}".
<patagonicus>
I personally always do LVM on LUKS, because that also makes it easy to move stuff and/or add other filesystems if needed, but it also gives me /dev/${volumeGroup}/${volume} paths for the actual filesystems.
<viric>
but I won't get the by-label before luksOpen
<viric>
I want a label to find what to luksOpen
<patagonicus>
Ah, misunderstood then. I usually go with by-uuid, For LUKS2 there's --label for luksFormat and config, which I think should be used for by-label.
<patagonicus>
You can also set up udev rules translating the UUID to a nicer name somewhere in /dev.
<viric>
ah I probably have luks2 everywhere, right?
<patagonicus>
Not sure when it was introduced, but I think any new luksFormat should use it. cryptsetup luksDump /dev/$foo should tell, without even needing a key.
<viric>
version 1.
<patagonicus>
:(
<viric>
meh.
<viric>
I thought luks 2 wasn't very new.
<patagonicus>
Did you create it with NixOS or something else? I wouldn't be surprised if e.g. Debian defaults to version 1 still. My current system is on version 2. Don't quite remember when I set it up, but probably some time in 2020 with the nixos installer.
<patagonicus>
Oh, there's cryptsetup convert $dev --type luks2. If you have backups.
<viric>
with nixos, sure
<viric>
patagonicus: they claim that a header backup should be enough
<viric>
worth trying! labels per luks partition would help me a lot
<viric>
thank you!
<viric>
patagonicus: all worked great!
<patagonicus>
Nice. :)
<viric>
now I only need a paper label for every hd and all will make sense with a simple script
<patagonicus>
I once made some elaborate QR code labels that encoded the UUID and also had the human readable name embedded in the QR code. Not really sure why because I could have solved my problem with some post its and a pen.
<eyJhb>
patagonicus: Post its do not like heat, and will quickly not stick anymore :p
<eyJhb>
But just a normal label will do
<patagonicus>
I guess, but the drives were cold storage sort-of backups, so it didn't really matter. And it was just a quicker way of picking the right drive out of four, worst case I could always check the serial number printed on them or just try.
<viric>
patagonicus: paper sticks and a pencil, here.
cole-h has joined #nixos-chat
waleee-cl has joined #nixos-chat
FRidh has quit [Quit: Konversation terminated!]
cole-h has quit [Ping timeout: 272 seconds]
srk has quit [Read error: Connection reset by peer]
srk has joined #nixos-chat
ece3 has joined #nixos-chat
tilpner has quit [Remote host closed the connection]
<infinisil>
E.g. if you have a 16-bit unsigned integer, 60000 < 1 with serial number arithmetic, because it takes into account that most likely the 1 was wrapped around the end
<viric>
interesting
rajivr has quit [Quit: Connection closed for inactivity]
<sphalerite>
why is firefox so bad at dealing with bad connections :(
<bbigras>
^ isn't it just the protocol and scp will use sftp now or something like that?
<clever>
bbigras: the scp binary on the CLI, still uses the old scp protocol, and all of its bugs
<infinisil>
clever: (not sure why you mention that?)
<clever>
just something interesting that i recently discovered
<bbigras>
clever: and it will still use it or it will switch to the sftp one?
<clever>
bbigras: according to the link i pasted, you have to choose to use the sftp binary, and learn its new cli args, or switch to rsync
<bbigras>
clever: oh. thanks
<clever>
they list several security exploits in scp
<clever>
for example, when you `scp host:foo.jpg .`, the server can just send you .bashrc instead, and the local client would happily overwrite .bashrc for you
<clever>
and when you run `scp foo.jpg host:bar`, that tells the remote sshd to just run `scp -t bar` and link the stdio up
<bbigras>
"As it turns out, Jakub Jelen is working on such a thing; it is an scp command that uses the sftp protocol under the hood. At this point, it is claimed to work for most basic usage scenarios; some options (such as -3, which copies files between two remote hosts by way of the local machine) are not supported. "Features" like backtick expansion will also not be supported, even though some users evidently think that this
<bbigras>
expansion might have legitimate uses. "
<clever>
`scp some-local-file remote:'`touch you-lose`remote-file'` that execution is done via sh, so it can be exploited with backticks to run commands rather then copying
<clever>
bbigras: that sounds like a replacement scp client, to fix the bugs by just using sftp for you
<bbigras>
yeah. and some options wont work so some script will break.
<clever>
it sounds like sftp has an interactive mode, like the old ftp clients, so that would have a lower learning curve for me, vs rsync
<bbigras>
you might be fine with the scp replacement. I wonder if it's an official project or just some guy on the openssh mailing list.
<clever>
and in the case of nixos, there is the tricky problem of ensuring the right scp has priority in all places
<__monty__>
clever: Fwiw, I basically always use the same flags for rsync and I get tab-completion of remote paths. Not a steep learning curve imo.
<clever>
__monty__: i tend to turn off tab completion, its often dumb and refuse to complete valid things, and it messes with my muscle memory by being too smart
<clever>
i can also see rsync+ssh remote completion causing a dozen pw prompts, if keys+agent arent working
<clever>
which reminds me, a recent nixos update broke gpg-agent
<clever>
ssh-agent now auto-starts, despite being disabled
<AMG>
damn, my thumbs are hurting from swiping on tinder!
<sphalerite>
amg: use the web app, so you can press keys on the keyboard instead?
<lunc>
lol
<__monty__>
clever: I get a password prompt for every completion but that's a choice ofc. Imo path completion is invaluable, especially when it's remote paths.
<clever>
__monty__: i just have an ssh open on the remote machine already, and confirm the path with things like ls
<clever>
or memorize the important ones, lol
<__monty__>
Maybe croc or magic-wormhole suits your needs?
<clever>
most of the scp'ing happens within the same LAN, where nfs could also be used
<__monty__>
clever: I used to do that btw, ssh and a bunch of ls. Never looked back after finding out the tab-completion : )
<__monty__>
It really is so much faster than using the clipboard to copy paths and such.
<sphalerite>
ControlMaster=auto and ControlPersist=300 ftw
<clever>
sphalerite: i always connect to machines with `loginto c2d` in the xfce run dialog
<clever>
and i also bound xfce's run dialog to an unused sun key, "again"
<__monty__>
sphalerite: I assume those are agent settings?
<sphalerite>
__monty__: no, ssh config settings
<sphalerite>
__monty__: it makes ssh reuse connections, so you only have to authenticate once
<__monty__>
What black magic is this?
<clever>
__monty__: the ssh client connects to another local ssh client, that is already connected to the same machine
cosimone has joined #nixos-chat
<clever>
__monty__: the ssh protocol already allows opening multiple tty's on one session
<__monty__>
Hmm, how does this interact with unstable connections? The kind that have ssh freeze up until a long timeout passes.
<sphalerite>
yeah that's the unfortunate bit, new attempts at ocnnecting will hang
<sphalerite>
__monty__: see also ServerAliveInterval for reducing that long timeout though.
<__monty__>
Hmm, think I'd opt for an agent instead. I'll keep it in mind though, thanks : )
<sphalerite>
I use both, with gpg-agent as the ssh agent, so I can use my yubikey to authenticate
<sphalerite>
I have to touch the yubikey to permit an authentication, so even though I don't have to type in the password it's very useful to not have to authenticate each time
<ashkitten>
amg: i stopped watching a while back because i got sick of him trying to chase perfection while ignoring fail-safes. if one part on that machine fails it'll probably ruin his whole concert, rather than missing a note or two...
<AMG>
ashkitten yup, he have battle tested each part of the machine, even fine tuning vibrations that caused a few marbles to drop on the floor lol
<ashkitten>
i'm fine with that, but he should be designing the machine to fail safely if a marble gets stuck. it's frustrating to watch him tune this machine so much when he's ignoring that if a marble gets stuck it could jam or break the entire machine instead of dropping on the floor
<ashkitten>
if it drops a few marbles during a concert that's preferable to it breaking entirely
<ashkitten>
but he hasn't paid any attention to that
<lunc>
That machine must be a nightmare to repair tho
<lunc>
I’d have nightmares about detecting which part had failed lol
<lunc>
As I doubt he has any sensors. It’s a huge, purely mechanical device. Like 2% electronics involved? Hehe
<samueldr>
do an apple, change the "whole mainboard" when something break
<samueldr>
I don't know how this translates, but something like re-do the whole machine?
<lunc>
anyone playing walking simulator here?
tilpner has quit [Remote host closed the connection]