Boosted by @oxy@social.bsdlab.au
I have so many issues with #unbound on my laptop while travelling that I'm tempted to port #unwind from #openbsd to #alpinelinux.
Please help, I already have too many projects.
Remote
Hugo 雨果
@whynothugo@fosstodon.org
mastodon 4.7.3Techsmith, interested in #SelfHosting, #OpenSource, #Sustainability, #DigitalRights.
#AlpineLinux Developer, #Nura TC and #OpenBSD enjoyer.
xmpp://hugo@whynothugo.nl or #whynothugo on irc.libera.chat
588 Followers
439 Following
50 Posts
Joined May 31, 2021
Website:
Sponsor:
Projects:
Co-Authored-By: GNU sed 4.9
Everything is always broken. A rant.
I went to fix a bug where a program of mine didn't properly render inactive border when unfocused. It turned out to be a compositor bug. Hours later: sent a patch.
I want to read web articles as epub on my ereader. I wrote a program to convert pages to epub. Now just need to render it. No program on Linux can render AND reflow epub files.
The obvious choice is to just write my own...
Replying to
@mttaggart@infosec.exchange The whole idea of a single central repository is absolutely terrible. Crates should be able to depend on crates hosted anywhere.
The combination of "cannot depend on anything on another hosting platform" and "hosting platform mediated by MS" is an awful combination.
With #neovim 0.12, the nvim-treesitter plugin is no longer required for tree-sitter usage. In fact, it's no longer maintained and archived. Support is built into neovim. Custom tree-sitter parsers can simply by installed by your favourite package manager, e.g.: `doas apk add tree-sitter-python`.
Summarised different approaches feasible for creating an XDG_RUNTIME_DIR: https://whynothugo.nl/journal/2026/04/02/creating-an-xdg_runtime_dir/
Short reflection looking back at pacman now that it's not part of my daily workflow: https://whynothugo.nl/journal/2026/03/26/pacman-syndrome/
Replying to
@codinghorror@infosec.exchange @bovaz@misskey.social @mjg59@nondeterministic.computer @jwz@mastodon.social It may not be perfect, but nothing else took its place, we don't really have anything that's better. Our options are RSS or simply don't follow any feed.
Must resist the temptation to write a clone of Notepad (the old-school version) for Wayland.
Replying to
@Strandjunker@mstdn.social Advanced civilisations collapse in the same way that wise old men die. It's not only plausible, it's the natural order of things.
Replying to
@witcher For online payments, the current Dutch standard (iDEAL) supports paying online without Visa/Mastercard as intermediary, so Wero seems like a strong step down.
And using a phone number as an account identifier is so absurd. As if phone number didn't expire or get re-used. We already have IBAN numbers which might be longer, but everybody copy-pastes them anyway.
So far they communicate that an app will be necessary and the only choice.
Malus: Clean room re-implementations as a service: https://malus.sh/
So real.
@niccolove@fosstodon.org Re: https://www.youtube.com/watch?v=WlgrCqgnk-M
Pretty neat and promising presentation, looking forward to seeing what comes out of this endeavour!
PLASMA STUDIO DEVLOG #0
Replying to
@ska@social.treehouse.systems I don't think it's a conspiracy to get you to link libsystemd. I think they're just honestly incapable of good API design, their hacky design turned out to be racey, so they added another ugly hack on top to work around it.
Why they wouldn't just pass a file descriptor is absolutely unfathomable.
Race conditions are all over the place, and systemd is full of hacks to work around them.
Replying to
Replying to
The README says that one of the motivations for this markdown parser was the need for a parser that "preserves source position of nodes".
But the parser doesn't preserve source positions, in favour of performance? 🤪
Delivery has failed to these recipients or groups:
notifications@github.com
The recipient's mailbox is full and can't accept messages now.
With technology being omnipresent, digital rights have become essential rights. The debate is no longer whether the government can install cameras in your house and cars. The hot topic is about how the private sector gets away with it. Eventually, in the age of computing, opting out of the all-seeing technocracy won't be an option.
Replying to
@ska@social.treehouse.systems @navi@social.vlhl.dev @hipsterelectron@circumstances.run I'll take the advise and use a lock before renaming, but I'll do so begrudgingly convinced that the OS should offer primitives such that this is handled at a lower level (e.g.: like atomic CAS).
(edit: files are never modified in-place, so that's not an issue)
Got some spam issues on GitHub. Reporting it requires solving a captcha which took 7 attempts. After that, GitHub form submission just fails.
Replying to
@ireneista I use $EDITOR only if I'm going to write additional lines. But usually if a project has a lot of working code for its "Initial commit", I'll be squashing lots of "ok so far" commits.
"Always on top" is such a terrible name of a feature that keep a window always in front.
Replying to
Replying to
@untitaker Online payments via iDEAL are ubiquitous in the Netherlands. Most stores don't even take Visa/MasterCard, and regular bank accounts don't include a card that's usable online either (you can get one, but it costs ~10x that of a regular bank card).
The risk of de-banking is still there: Google or Apple can block your account and your device along with it. There's precedent of people not being able to use apps any more because their account was banned.
Replying to
@untitaker Windows? They could just target browser. The current Dutch online payment system supports paying via a browser (a few banks don't), and the web is ultimately platform-agnostic.
Replying to
@witcher To be clear: this isn't _replacing_ US services in the payment area, it's just changing which ones we rely on. From Visa+Mastercard to iOS+Android.
I don't find this to be an improvement. Someone can use Maestro card (the usual in NL) without having to buy a device from Google/Apple. This just serves to erode the influence of phone vendors done outside the US.
Replying to
@alerque I use ctrl+r to fuzzy-search in history, so I use, for example, "ctrl+r,am" to match "git commit --amend".
I have a few aliases for git, s=git status --short and p=git push -u origin HEAD, f=git fetch.
Replying to
@ska@social.treehouse.systems @navi@social.vlhl.dev @hipsterelectron@circumstances.run
I find locks to be a pain for this kind of thing, you want atomicity without such high level locks. Bubbling a lock this high up makes operations as slow as possible, while relaying all the complexity all the way.
"Avoiding concurrent access" is unrealistic for user files. E.g.: calendar data. A sync tool might want to operate at the same time as a user making an edit.
Replying to
@elly I'd not heard about this before. Does it have a lesser crash than the "regular" (non-CR) variant?
Replying to
@ska@social.treehouse.systems @navi@social.vlhl.dev @hipsterelectron@circumstances.run What I'm asking for is basically a way to implement a "compare and swap" atomic operation.
Why is it that sometimes I can't run `doas eject /dev/sdb` (I get `eject: unable to eject`) on an SD card? I see the light blinking on it, so I want a synchronous command when I can be CERTAIN that it's done writing and flushing.
"How hard could it be?"
Common words spoken before diving into a 5 minute task or a 5 month nightmare of pure pain.
Replying to
@navi@social.vlhl.dev @hipsterelectron@circumstances.run @ska@social.treehouse.systems Why would readers use locks? That doesn't seem to solve any issue in such contexts.
Replying to
@navi@social.vlhl.dev @hipsterelectron@circumstances.run @ska@social.treehouse.systems
Right, the usage of a FD is the key point: I don't mean for the contents to be compared, just whether the FD points to the file.
The only reason concurrent workers are "a very bad idea" is because we lack this kind of CAS semantics. The only reason we need things like locks is that too. They're workaround, patches, that we've normalised at "the right way" because we don't have better filesystem semantics.
Replying to
@cadey@pony.social Your blog has some pretty good posts, but these garbage posts showing up all the time are a real annoyance.
Replying to
@smlavine When thinking of such metaphors: what do you call surfaces which are on the top or the bottom of the screen? I've always seen them referred to as "top" and "bottom".
Replying to
@wishy iDEAL is pretty nice and simple (and ubiquitous within NL), it's a shame that it never expanded beyond the Netherlands.
Not sure if a troll, or just a very confused soul.
Replying to
@ska@social.treehouse.systems @navi@social.vlhl.dev @hipsterelectron@circumstances.run
Take this sequence:
- Read file
- Determine that we want to replace it
- Write content into a separate file, preparing the replace the original
- (race, some other process has replaced the file!)
- Replace the file
This cannot be prevented with renameat2. If nothing goes wrong, you can detect the problem (it's already too late tho). If your luck continues, you can revert it, and pray that no other process has read the bogus file in the meantime.
Replying to
@smlavine We also have some asymmetry in that some windows are in the background or behind others, but we then call the ones in front "on top"?
I don't know where the expression comes from, but its ambiguity bothers me greatly.
How would you connect wires to this UART (GND/TX/RX) if you don't have a solder at hand?
Replying to
@ska@social.treehouse.systems @navi@social.vlhl.dev @hipsterelectron@circumstances.run
I know what locks are and how they work, I don't see how readers taking a lock would have any relevant impact when replacing a file via rename or renameat.
Replying to
@untitaker "Fully European" and "Usable only with a device and a contractual agreement with either Google or Apple" don't sound consistent.
The state of the art in the Netherlands lets me pay online to local online merchants without any American company involved, using a regular browser.
An interesting aspect of git-based projects is that you can work on them studying commits and understanding history within a given scope, without really having to comprehend all rest of the codebase.
I.e.: I can have a _temporal_ understanding of a portion of the codebase (how it progressed over time and the constraints of those changes) rather than a _spatial_ understanding (how it interacts with its surroundings).
Replying to
@navi@social.vlhl.dev @hipsterelectron@circumstances.run @ska@social.treehouse.systems I never said anything about modifying a file tho. I want to replace the file. Modifying in-place means that readers can read a partially written file, and a system outage permanently corrupts the file to a half-written state.