<colemickens>
do the serokell folks hang out somewhere? I want to know more about their plans to utilize systemd portable services.
Rainy_Day has quit [Remote host closed the connection]
<abathur>
unrelated: a bot that searches GH for repos with strongly-worded please-don't-use-this statements and opens issues that mention how much you're enjoying using it for some project in <important sounding lab/facility/data center> and then ask some perplexingly indecipherable question?
das_j has quit [Quit: Bridge terminating on SIGTERM]
<LinuxHackerman>
I remember back in the day when it was oHg
<LinuxHackerman>
Am I right in believing that linux handles files becoming unavailable without warning pretty poorly (i.e. processes hang forever while waiting for IO that will never complete, even if it can be determined that the IO definitely won't complete)?
<LinuxHackerman>
Can anyone ELI5 why?
<DigitalKiwi>
is that not the halting problem
<sphalerite>
no
<sphalerite>
it's a question of IO
<sphalerite>
like if a file has ceased to exist due to a filesystem rollback, that file is not coming back
<DigitalKiwi>
maybe it is? ;P
<sphalerite>
sure, you might end up with a file that has the same name and even the same contents. But it's not the same file
<DigitalKiwi>
what is a file if not its name and contents
waleee-cl has quit [Quit: Connection closed for inactivity]
<DigitalKiwi>
for that matter even the name isn't really that important
<sphalerite>
echo "hello" | tee a b
<sphalerite>
would you say that after this, a and b are the same file?
<DigitalKiwi>
kiwi@mvp-nixos ~ []$ diff <(echo "i'm a file *wink*") <(echo "i'm a file *wink*")
<DigitalKiwi>
kiwi@mvp-nixos ~ []$
<sphalerite>
right, but diff compares the contents of files.
<DigitalKiwi>
07:00 sphalerite: would you say that after this, a and b are the same file?
<sphalerite>
And I'm talking about regular files here, not pipes. Regular files are mutable objects, and even if two files have the same contents that doesn't mean they're the same file.
<sphalerite>
Since mutating one doesn't affect the other.
<DigitalKiwi>
yes (zfs dedup)
<sphalerite>
zfs dedup is an optimisation, not something that changes the semantics of files.
<DigitalKiwi>
two files are the same until they are different and then you have two files
<sphalerite>
And I'm not really here for a philosophical discussion but a practical one
<sphalerite>
for the purpose of my question, the file is what the process that has it open has a handle on
<sphalerite>
and after a rollback, that handle will never become valid again
<DigitalKiwi>
whether or not that matters depends on what the program is doing though
<sphalerite>
It matters as soon as the program tries to read from it
<DigitalKiwi>
if you're network goes away and your nfs disconnects when it comes back it's the same nfs
<sphalerite>
and I think it's kind of awful that what happens then is that the process hangs in an unkillable state
<DigitalKiwi>
i'm not disagreeing that it is not handled poorly btw but i'm not sure that it could be handled "better" because what is better is entirely context dependent
<sphalerite>
I'm pretty sure it could be handled better? I/O errors for the ZFS rollback, _some_ way to tell the kernel "yeah that nfs isn't coming back, unhang the process."
<sphalerite>
I don't like that the only way to "fix" some of these things is a reboot
<DigitalKiwi>
just like how my computer is now 10x slower because of mitigations to fix exploits that are a problem if you're running vms on untrusted hardware at amazon but i am not amazon
cole-h has quit [Ping timeout: 245 seconds]
<sphalerite>
Those can be disabled, no?
<DigitalKiwi>
i should find out how if they can lol because i have had several hundreds of lockups in the past couple of months lol
<DigitalKiwi>
and i phrased that poorly i don't mean that things can't be improved at all just it's hard problem to solve especially in a generic way
<sphalerite>
DigitalKiwi: boot.kernelParams = ["mitigations=off"]; will disable them, but keep in mind that some of the mitigated vulnerabilities can be exploited via javascript running in your browser.
Ashy has joined #nixos-chat
Ashy has left #nixos-chat ["WeeChat 3.0"]
lucus16 has quit [Quit: WeeChat 3.0]
spacekookie has quit [Quit: **aggressive swooshing**]
<patagonicus>
Hmm. I have an old GitLab install on Docker and I need to figure out how to upgrade it. I also want to move it out of Docker into NixOS. I feel like setting up a VM on an old version of NixOS might actually the best way - move it out first, then upgrade using NixOS.
<patagonicus>
Extra fun bit: it's using MySQL and I have to migrate it to Postgresql because GitLab dropped support for MySQL.
<patagonicus>
Extra extra fun: GitLab fails to make backups. :/
nckx is now known as jorts
cole-h has quit [Ping timeout: 245 seconds]
aei has joined #nixos-chat
omnd has quit [Ping timeout: 265 seconds]
waleee-cl has joined #nixos-chat
jorts is now known as nckx
rajivr has quit [Quit: Connection closed for inactivity]
sterni has quit [Ping timeout: 272 seconds]
sterni has joined #nixos-chat
j is now known as jess
<__monty__>
Hmm, I would like to get the date of the start of the week. `date -d 'last monday'` gets close but is wrong every monday. Any way to specify "last monday before tomorrow?"
<__monty__>
Ah, date -d 'last sunday + 1 day'` seems to do the trick.
<colemickens>
those Mfr/Product/SerialNumber look questionable.
<colemickens>
Alright, it's time to remove this DP adapter and see what happens...
<colemickens>
with the DP adapter removed, the device name has returned to normal.
<colemickens>
I feel like I'm being dramatic or something but I can't help feel like there might be some sort of insidious bug I just ran across
<samueldr>
oh, don't worry
<samueldr>
it might be on any of the parts in the path
<colemickens>
:)
<samueldr>
(yes worry)
<samueldr>
especially since, IIRC for thunderbolt there is literally two implementations, both from the same vendor
<samueldr>
and, memory is fuzzy about the details, it's not possible to ship thunderbolt from another vendor becaus of... ????
<colemickens>
lol, after a reboot: Bus 003 Device 013: ID 2c97:1015 Ledger USB 2.0 Hub
<colemickens>
time to try a different chinese usb3 hub
<colemickens>
but the DP port wasn't plugged into the hub, making me think the bug is "further up the chain"
<colemickens>
but who knows it could just be the trigger. lord what a nightmare.
<colemickens>
next laptop will have usb-a ports
<colemickens>
(can't believe I became one of those guys :P)
<samueldr>
I'm actually quite concerned about usb and extensions from type-c
<samueldr>
things seem bad all around :/
<samueldr>
especially thunderbolt is making me squirm
<samueldr>
I'm actually taking in consideration whether a computer supports thunderbolt or not when looking
<samueldr>
because... oof...
<colemickens>
lmao, different hub, linux actually loops on setting them up and tearing them down now.
<colemickens>
lots of attempt power cycle, etc. still cycling....
<samueldr>
even standard usb A with usb 3 makes me worry sometimes
<samueldr>
I have a setup with a specific device which ureliably makes the whole usb stack keel over
<samueldr>
the usb stack on the host
<colemickens>
then, we get this pcieport BAR failures and then USB basically stops working until I reboot. woohoo I've achieved usb nirvana again. definitely nothing wrong here.
<colemickens>
maybe I could avoid USB by leaning on BT, that sounds fool proof, right
<samueldr>
IIRC the naming also folds in "high speed, full speed and low speed", what are usually referred to as "2.0", "1.1", and "1.0" (order may not be right)
<samueldr>
usb 4, though, has DMA because USB 4 is "SuperSpeed+ and Thunderbolt 3"