Wolf480pl

Sysadmin stuck with k8s,
Linux nerd
Likes The Unix Way
🇬🇧🇵🇱(🇯🇵 a bit)

Public Key

npub170403ynslsp4tn32j8tfhjn3cdq2pgut0jpnhrlcxl524gxf36pqput88d

NIP-05 Address

wolf480pl@mstdn-io.mostr.pub

Profile Code

nprofile1qqs086hcjfc0cq64ec4fr45mefcuxs9q5w9heqem3lur06925rycaqspz4mhxue69uhhyetvv9ujuerfw36x7tnsw43qzm7vk2

Publishing to

relay.ditto.pub

Author Public Key

npub170403ynslsp4tn32j8tfhjn3cdq2pgut0jpnhrlcxl524gxf36pqput88d

Show more details

Last Notes

2026-05-17 19:03:23 UTC

- reply

@nprofile…6srv why in every log message instead of one at the start of the log? Because the start of the log may've been days ago, rotated somewhere else, and also people tend to send around individual log messages when talking about problems.

2026-05-17 17:34:09 UTC

can I just
void foo(const char *arg0, ..., /* , (char*) NULL */) {
const char **argv = &arg0;
/* ... */
execvp(argv[0], argv);
}
or is that UB?

2026-05-13 10:13:40 UTC

- reply

@nprofile…eqww
I'm guessing this:
union {
char bytes[3];
struct __attribute__((packed)) {
int lo:12;
int hi:12;
};
};
doesn't do the right thing?

2026-05-10 11:13:09 UTC

- reply

@nprofile…hcxx
yup
relocation R_X86_64_PC32 against absolute symbol `memb1' in section `.text.startup' is disallowed
it would've worked on i386 right? cause that one doesn't have IP-relative LEA?

2026-05-06 19:02:33 UTC

Have modern programming languages given up on linking?

2026-05-04 19:36:39 UTC

- reply

@nprofile…hcxx
https://www.youtube.com/watch?v=rStL7niR7gs

2026-05-02 10:43:24 UTC

- reply

@nprofile…tgjm
> Space Shuttle
some things have changed a lot
> Tomahawk
others have not

2026-04-28 19:50:14 UTC

- reply

@nprofile…gpjx writing retro-scifi books for other burned out tech workers

2026-04-26 18:45:06 UTC

- reply

@nprofile…hcxx
ext4

2026-04-26 17:54:33 UTC

- reply

@nprofile…hcxx
(used rate(...[1h]) and assumed it worked like increase(...[1h]))
after fixing the query, it's:
~19 GiB written that node-exporter saw during the last 8 months

2026-04-26 17:52:19 UTC

- reply

@nprofile…hcxx wait no I calculated it wrong

2026-04-26 17:41:49 UTC

- reply

Actually, less than a year - this one lasted 8 months.
What am I doing wrong?

2026-04-26 17:37:13 UTC

Why do microSD cards develop ioerrors after a year in an.SBC...

2026-04-22 21:50:25 UTC

Random reflection:
The most valuable information about a piece of code, or a patch, is not how it works. Or why it works. Or who wrote it.
The most valuable information is why it was done in that way in particular.
Did someone implement it the first way they came up with, without giving it much more thought?
Did the author know a better way, but was in a hurry so knowingly used a quick workaround?
Or is there a very subtle reason that this has to be done this way that took days to discover?

2026-04-22 10:20:52 UTC

- reply

@nprofile…22w0
counter-intuitive fact: when Buddy Holly was Paul McCartney's age, Paul McCartney was *not* Paul McCartney's age.

2026-04-21 10:27:25 UTC

- reply

What if I have a chunk of memory with a bunch of counters that are being constantly updated by another thread, and I want to make a private copy of it (doesn't have to be consistent as a whole, only each counter needs to be internally consistent) to read at my leisure, without the other cores' writes pushing it out of my cache?
Not a real case where I found this to be a bottleneck, just wondering what are the possibilities.

2026-04-21 10:24:59 UTC

If you tell a CPU to load a word from memory into a register, and then proceed to not use that register for a few tens of instructions, it'll run those instructions before the load has finished, right?
But what if he thing you want to prefetch does not fit in a register?
1/

2026-04-19 09:13:44 UTC

- reply

@nprofile…22w0
kinda sad that all the cool things were already discovered in the 60s by IBM and no cool things were discovered in the 2010s

2026-04-16 21:51:03 UTC

Should I update BIOS on an ASUS B650 motherboard if current BIOS seems to work fine but is 2 years old... and there are several versions in between with "after installing this, you cannot downgrade" ?

2026-04-16 09:38:21 UTC

- reply

@nprofile…s997
I admit a large part of it is min-maxing for fun
@nprofile…9eh0

2026-04-16 09:37:18 UTC

- reply

@nprofile…s997
well, more like "would like to, if possible without buying another cooler"
@nprofile…9eh0

2026-04-16 09:35:53 UTC

- reply

@nprofile…s997
I want to have a margin for 36°C ambient in the summer. Currently it's 21°C ambient.
@nprofile…9eh0

2026-04-16 09:29:51 UTC

- reply

@nprofile…9eh0
Pretty sure the 3D-vcache is on top of the die that has cores, but yeah this one is off to the side, and the IO die (memory controller + pcie + misc IO) is in the center.
But yeah, I've heard if offset cooler mounts. If it turns out this OC struggles in the summer and I end up buying a better cooler, I'll look into those.

2026-04-14 22:30:31 UTC

- reply

@nprofile…z99u cause like
I wouldn't expect companies to deal with double standards
they create double standards, and other people have to deal with them (or give up), right?

2026-04-14 22:22:59 UTC

- reply

@nprofile…z99u I mean, do lawyers hired by companies and countries, and lawyers hired to litigate against companies and countries, have to deal with such double standards often?

2026-04-14 21:49:31 UTC

- reply

@nprofile…z99u
wait, so this kind of double standard isn't something lawyers deal with on a daily basis?

2026-04-14 21:35:55 UTC

- reply

@nprofile…z99u
so if they were to accidentally leak the code, they'd have no way of preventing further redistribution of that code?

2026-04-14 16:45:44 UTC

- reply

Also, is it normal that this CPU cooler - Thermalright Assassin Spirit 120 V2 - can't get the CPU below 86°C at 120W load? (21°C ambient)

2026-04-14 16:34:11 UTC

- reply

1.04V is the default VSOC btw
Somewhat relatedly, enabling EXPO adds 8W to the idle CPU power draw, and increases temps under full load (prime95 smallFFTs + furmark on iGPU) by like 5°C.
Even if I can't get UCLK to 3000MHz, surely it should be able to do 1500MHz at a way lower VSOC, since it's doing 2400MHz at 1.04V...
Hopefully that would lower the temps?

2026-04-13 16:55:12 UTC

- reply

@nprofile…hcxx I don't know how this happens but I tend to find bugs in most code I look at...

2026-04-13 16:30:58 UTC

- reply

@nprofile…hcxx but , if I'm reading the code correctly, it will delete :text and treat :text2 as a new snippet header

2026-04-12 14:08:18 UTC

why does rendering fur make GPUs exceptionally hot?

2026-04-06 18:00:12 UTC

- reply

@nprofile…8r3h @nprofile…j6pd
> [in the office] emergencies are rarely this immediate
Was about to comment the same.
It sounds like this advice would be great if you have to make a decision in seconds.
But in the office, where you usually can slow down to gain more time to evaluate your options, but management is overeager and wants you to go fast, I think that making them aware of the severity of a problem and reasons why the current approach won't work is half of the work...

2026-04-03 14:33:07 UTC

- reply

@nprofile…gd52 what I'm more concnerned about is anyone with (bought or enforced) access to Telegram servers bypassing the sticker verification and using the alleged RCE...

2026-04-03 14:21:02 UTC

- reply

@nprofile…gd52
AFAIU that validation is when uploading a new sticker, not when using an already-defined one.
> everything you say or send on Telegram goes trough their servers in plain text
where did they say that?
Also AFAIK they only ever claimed that the special "secret chat" mode that nobody uses is end-to-end encrypted, the regular chats were never claimed to be end-to-end encrypted, so obviously the server had access to it...
But that's neither news nor RCE.

2026-03-31 17:21:29 UTC

- reply

@nprofile…tpfq yeah it's crazy how at 7 PM it's still bright outside

2026-03-31 05:41:07 UTC

- reply

@nprofile…z99u
AV:L/AC:H/.../UI:R ?

2026-03-30 21:56:34 UTC

- reply

@nprofile…z99u more like, is not exploitable / is exploitable only by their servers

2026-03-28 18:47:17 UTC

> Instead of building the biggest thing in the smallest amount of time, I am now trying to spend the most amount of time building the smallest thing, and there's a lot of joy to be had in that
-- MumboJumbo, https://www.youtube.com/watch?v=_Nd5UsAPC5I&t=1422s
So much this.

2026-03-28 12:55:06 UTC

- reply

@nprofile…waqh
Yeah, I know that. But what you're describing requires the phone to be *turned on*, right?
And I'm explicitly asking about the statement from OP's article that says a *turned off* phone still communicates via bluetooth.

2026-03-28 11:39:51 UTC

- reply

@nprofile…xj4x
oh, most of these qoutes are from
https://www.wireguard.com/known-limitations/
and
https://www.wireguard.com/todo/
?
Hadn't looked into those, I guess I got confused by the "interworkings" menu name.
Hm... maybe a BFD daemon with an action script that adds/removes routes would be the right solution after all.
And if it was well integrated with OpenWrt, I imagine it'd be convenient in all kinds of situations...

2026-03-28 11:10:40 UTC

- reply

@nprofile…xj4x
If wireguard wants to be a primitive, a building block in larger solutions, that's fine, but:
- I couldn't find anywhere stating that
- Doing both routing and filtering based on AllowedIPs looks out of place. It's not suitable for any advanced usecase - you end up needing one interface per peer anyway. So it only makes sense as a way to make WG easier to use stand-alone.

2026-03-28 11:01:47 UTC

- reply

@nprofile…xj4x
from https://www.wireguard.com/ :
> WireGuard aims to be as easy to configure and deploy as SSH. A VPN connection is made simply by exchanging very simple public keys – exactly like exchanging SSH keys – and all the rest is transparently handled by WireGuard. [...] There is no need to manage connections, be concerned about state, manage daemons, or worry about what's under the hood.
idk to me that does kinda imply "you don't need other things on top"

2026-03-28 10:18:37 UTC

- reply

@nprofile…0u3c
yeah but does he? I can't find evidecne that he does, and wireguard website still has the banner of EdgeSecurity at the bottom...

2026-03-28 10:14:23 UTC

- reply

@nprofile…0u3c
does Jason work for Tailscale now? o.O

2026-03-28 10:08:22 UTC

- reply

Well, wireguard is not much different from GRE orIPIP in that regard, so a tunnel agnostic reachability detection daemon that adds and removes statically configured routes would've been a more universal solution. It could eg. use BFD.
Still, if wireguard is trying to be a "batteries included" type of solution, I'd expect it to have such a feature, at least in the userspace tooling part.

2026-03-28 10:01:03 UTC

Imagne wireguard but it detects if the peer is reachable and only adds routes when it is.

2026-03-28 09:12:44 UTC

- reply

@nprofile…e965
> even if your phone is turned off, it will still send out Bluetooth signals to broadcast its location
wtf? where can I read more about that?

2026-03-27 19:55:09 UTC

- reply

@nprofile…qvn2 hmm so it's a slightly weaker ThinkCentre M715q but in laptop form?

2026-03-27 18:14:47 UTC

- reply

@nprofile…lja2
> fully Microsoft-compatible
that's a very ambitions goal

2026-03-27 18:12:49 UTC

- reply

@nprofile…lja2 getting onlyoffice to feature parity with google docs would be amazing

2026-03-26 12:12:21 UTC

- reply

@nprofile…fccl
xDDD
I was thinking more like
force every building with a neon sign / big-ass backlit logo to put city name in it.
"IKEA WARSZAWA"
"ORLEN KRAKÓW"
"TOYOTA KIELCE"

2026-03-26 12:06:38 UTC

Is there an online marketplace for renting roof space for antennas, or is this a "cold-send an email asking for a quote" type of market?

2026-03-26 11:56:46 UTC

- reply

@nprofile…0u3c
pkin is only visible from within Śródmieście, otherwise it's either fog or buildings near you that instruct it.
Also, not every Random city has a landmark like that.

2026-03-26 11:46:17 UTC

Unix shells tend to show the local hostname in the prompt, so that you see it at all times.
Cities, OTOH, don't have their city name written all over the streets. It's usually written next to major roads when entering the city, but once you're inside, it may be difficult to find out which city you're in.

2026-03-25 10:54:24 UTC

- reply

@nprofile…lja2
I guess "telephone" implies POTS.
In that case I'm not surprised - phone networks are notoriously insecure.

2026-03-25 10:53:05 UTC

- reply

@nprofile…lja2
do we know if it was a POTS (plain old telephony service) call, or some app like Signal or Google Meet?

2026-03-24 17:06:24 UTC

- reply

@nprofile…tgjm
packet core? why not cell-switched?

2026-03-23 20:08:17 UTC

- reply

@nprofile…0u3c @nprofile…al9h
right?
Yet the article somehow makes it sound like it would've been less worthy of mockery.

2026-03-23 20:04:39 UTC

- reply

@nprofile…0u3c @nprofile…al9h
Yeah, but I don't get why getting damaged in combat would be somehow more virtuous.
And in general, I don't like the surface-level mockery - it seems to imply that US has to be utterly incompetent to let those things happen, which I disagree with.
I think the US is pushing its luck to seize a once-in-a-decade opportunity.
Now, if someone were to argue that there was no chance this would succeed, and US should've known it from the start, that'd be another story.

2026-03-23 19:48:36 UTC

- reply

@nprofile…0u3c @nprofile…al9h
AFAIK its scheduled maintenance time was in the middle of the Special Maduro Operation.
Also, the US did not exactly choose when to start a war - AFAIU they were in a rush to start it while there's still a potential for protests in Iran.
It seems they're 1 carrier short - if they had one more, they could've sent it to the Gulf to replace Ford, and be able to intervene when the protests were happening, instead of doing that 1.5 months later.

2026-03-23 19:37:30 UTC

- reply

@nprofile…al9h
The ship is going to port for maintenance *because* of those 266 consecutive days at sea.
It was long overdue, and when your maintenance is long overdue, shit starts to break. That happens to almost every machine. Nothing to make fun of.
If the author wants to make fun of sth, make fun of US not putting another carrier in the Gulf immediately after Ford left for Venezuela.

2026-03-23 19:32:08 UTC

- reply

@nprofile…al9h
> The USS Gerald R. Ford, meanwhile, the most expensive warship in human history, is retreating to Crete. The official reason is a “laundry fire”. 266 consecutive days at sea, 28 days short of the Vietnam-era deployment record, and the crown jewel of the US Navy is fleeing the theatre, not because of being damaged in combat, not because missiles are flying around it… But because someone's skivvies got too hot.
This is either ignorant, or written in bad faith.

2026-03-22 21:22:36 UTC

- reply

@nprofile…nhhd @nprofile…0x09
It's not a reimplementation of Unix tools.
It's an embrace-extend-extinguish attack on Unix tools.
It's a backwards-compatible extension of a reimplemenation of Unix tools.

2026-03-22 21:16:44 UTC

- reply

@nprofile…vnlz @nprofile…0x09
or by "it ain't very good" do you mean "the philosophy isn't good" ?
If so, then I think that's a non-sequitur.
Failure to adhere to a principle doesn't mean the principle isn't good.

2026-03-22 21:14:12 UTC

- reply

@nprofile…vnlz @nprofile…0x09
Yeah, all software sucks, news at 11.
Doesn't mean we shouldn't try to make it better.

2026-03-22 21:06:02 UTC

- reply

@nprofile…0x09
Now, if you consider following Unix Philosophy to be a yes-no matter, then what I'm saying might sound like "No True Scotsman".
But I think it's more of a direction.
An unattainable ideal that some software authors try to approximate.

2026-03-22 21:05:14 UTC

- reply

@nprofile…0x09
> Hell, consider GNU true which can return 1. Just run this! /bin/true --help >&- and see what's in $?. It had one job.
I don't think anyone would consider GNU coreutils (or any GNU software for that matter) a very good example of Unix Philosophy, and `/bin/true --help` was the top example of where it fails.

2026-03-21 23:48:29 UTC

- reply

@nprofile…fccl does that mean they're gonna sell the stocks to achieve the new split, or does the new split only apply to future purchases?

2026-03-21 20:54:17 UTC

- reply

@nprofile…c057 I do not understand it one bit, but I guess that's the charm of 2009...

2026-03-19 22:05:12 UTC

- reply

@nprofile…lja2
> propaganda is spreading online about a "Narva people's republic" in #Estonia.
> [...] War is coming to Europe.
Are you saying that now Russia is able to wage two wars at the same time?

2026-03-19 22:03:58 UTC

- reply

@nprofile…lja2
> the moment Russia wins or achieves an unjust peace, their next consideration will be [...] breaking that peace to conquer more of Ukraine
Ukraine will do its best to not agree on a peace deal that would allow Russia to do that.
In order to pull that off, Russia would need a exceptional information asymmetry, or pressure on Ukraine from the West greater than Trump's last attempt.
2/

2026-03-19 22:00:44 UTC

- reply

@nprofile…lja2
> If we allow Russia to win in Ukraine, or to achieve an unjust peace, it will be a matter of years before Russia attacks the European Union
Here's the trick: we don't.
> If we choose to wait, we are only giving Russia the time it needs to rearm and prepare to attack us
No, we're also giving ourselves time to rearm and prepare to defend.
1/

2026-03-19 20:45:13 UTC

- reply

@nprofile…gcvv
lizard have the ability to dispose of their tails

2026-03-19 20:35:16 UTC

- reply

@nprofile…gcvv
disposable?

2026-03-19 13:11:07 UTC

- reply

@nprofile…lja2
did you know about it when you posted https://eupolicy.social/@jmaris/115939940687023167 ?

2026-03-18 22:44:41 UTC

- reply

@nprofile…gcvv are you linking to a regression or a bugfix?

2026-03-09 07:52:35 UTC

- reply

@nprofile…z99u you just made me realize I never use capslock...

2026-03-08 12:43:08 UTC

- reply

@nprofile…hhq5 if there was a way to disable travis on the repo without removing .travis.yml, I would keep the file "for now", in case the new CI turns out to have issues
then when the new CI doesn't have issues, I'd forget about it

2026-03-08 12:18:38 UTC

- reply

@nprofile…hhq5
why delete a file that just sits there, out of your way

2026-03-08 09:43:06 UTC

- reply

@nprofile…hhq5
so they're a capable enemy?

2026-03-03 19:09:25 UTC

- reply

@nprofile…fccl
what's more appalling than a corporate PR stock photo?
corporate PR AI-generated image

2026-03-01 22:19:10 UTC

- reply

@nprofile…vpvm oh, but clang now does the same annoying thing as gcc and uses node->next when they are equal... sigh... https://godbolt.org/z/T3dfKfcr6

2026-03-01 22:16:53 UTC

- reply

@nprofile…vpvm I'm amazed that if does not compile to a cmovne

2026-02-28 15:11:53 UTC

- reply

@nprofile…lja2
> the step short-circuited its formal approval by EU lawmakers.
No matter how good a deal it is, the executive bypassing the established processes is bad.
The same kind of bad as Trump ignoring US Congress and court orders.
The same kind of bad as EC ignoring ECJ orders regarding Chat Control.

2026-02-27 18:31:54 UTC

- reply

@nprofile…nhhd btw. I wonder how much stuff actually uses diffserv

2026-02-27 18:31:14 UTC

- reply

@nprofile…nhhd
As for per-program balancing - tc-flow classifier can hash locally-originated traffic by socket uid and gid, but sadly not by pid. Also, traffic class can be set using cgroups (see tc-cgroup) but that doesn't sound very useful either.
Unless you use flatpak, I guess.
In any case, the place you want to do traffic shaping is right befeore the bottleneck link, which in most cases is not the link from your computer to the home router...

2026-02-27 18:30:21 UTC

- reply

@nprofile…nhhd
IIRC torrent clients are only supposed to unchoke four peers at a time... but that might be 4 peers per torrent, so if you have many, yeah, it could get bad.

2026-02-27 18:12:15 UTC

- reply

@nprofile…nhhd I'm not saying that whoever wrote the code to make 10 connections intended to game TCP fairness.
But a side effect of opening 10 connections is an unfair advantage in TCP congestion control & traffic shaping algorithms.

2026-02-27 17:07:27 UTC

- reply

Jim Gettys in "Traditional AQM is not enough"[1] complains, among other things, that web browsers game TCP fairness by opening 10 or more connections
Later in that article, and also in other articles on that topic, it is claimed that fq_codel is a solution, because it separates flows and gives each one a separate queue
I kinda expected they found a way to treat all TCP connections from one webpage loading as the same flow, but they never said anything about it
2/
[1]: https://gettys.wordpress.com/2013/07/10/low-latency-requires-smart-queuing-traditional-aqm-is-not-enough/

2026-02-27 17:06:18 UTC

I've been reading about bufferbloat, traffic shaping, fq_codel, and stuff, and there's one thing that kept puzzling me:
How do they classify flows?
1/

2026-02-27 16:23:54 UTC

- reply

@nprofile…re7n @nprofile…sse5
do spreadsheets allow for circular dependencies between formulas?

2026-02-27 16:02:14 UTC

- reply

@nprofile…sse5 @nprofile…re7n
This was already mentioned in other replies.
But also, I'm not claiming that I am noticing LLMs being intelligent. I am not.
I'm just saying that OP's argument is insufficient to prove that no LLM constructed in this manner will ever be intelligent.

2026-02-27 15:00:57 UTC

- reply

@nprofile…hhq5 all this talk about feedback loops and yet the dog neither sees the game, nor is rewarded for the game being good? Well, so much for LLMs being "tools"...

2026-02-26 21:27:07 UTC

- reply

@nprofile…re7n pretty sure that's a fallacy, kinda like "a sculpture is just stone, therefore it can't be beautiful", or "a cell is just a bunch of proteins, therefore it cannot be a living creature".
Now, I'm not saying a huge database of probabilities can be intelligent (I hope it can't), just that I think a better argument is needed why in this case, being a database of probabilities prevents it from being intelligent.

2026-02-26 20:27:56 UTC

- reply

@nprofile…re7n on the flipside, I feel like some people use the term "stochastic parrot" or "it just completes the next token" to imply that "therefore it cannot be intelligent" - is that correct reasoning?

2026-02-26 01:17:37 UTC

- reply

@nprofile…nhhd we have an active phased array at home:
https://media.mstdn.io/mstdn-media/media_attachments/files/116/134/284/960/563/076/original/1d13b54a63f14309.jpg

2026-02-22 19:17:12 UTC

- reply

@nprofile…lja2
> life became more and more expensive, housing is out of reach for most young people
do you have data to support that?
I've heard this repeated many times, and it kinda feels that way, but the only person I saw dig for numbers to verify that claim concluded that the standard of living of 20yos these days is higher than of our parents or grandparents when they were in their 20s.
Unfortunately I don't have the link at hand, I'll see if I can find it...

2026-02-22 14:42:19 UTC

- reply

@nprofile…gcvv TLS (Transport Layer Spectrometer)

2026-02-21 16:30:00 UTC

- reply

@nprofile…hcxx
I thought ffmpeg had its own and it was faster, but maybe that's a recent development.