So apparently firefox released a new ESR version. And at some point in the intervening versions, "fixed" a longstanding "hover tooltips persist after window loses focus" bug. With a fucking sledgehammer. The 'solution' was to completely disable hover tooltips for unfocused windows. And then marked the issue for "uh, people have multiple screens and that now requires an extra click to see hover tooltips, then click back to what they were doing" as a WONTFIX "because none is better than stuck."
Despite the fact that this change breaks over 100 versions of behavior, and also makes firefox work differently than literally every application I know of that uses hover tooltips, and will happily display them when unfocused.
(they also added an annoying dimming affect to the tab bar without an off switch, but that's at least fixable via CSS... Still shows that they give no shits about multiple screens being a thing...)
=> More informations about this toot | View the thread
Stupid idea: "Kerberos HSM" for machine credentials and service accounts. Keep the machine credentials locked up inside it, and the machine can use them, but can't extract them. Any HSM that supports the right algos with "raw" operations should work, but a specialized one can potentially handle things more "efficiently"
Also makes it possible to securely distribute service credentials to end machines, without allowing a compromised machine to exfiltrate them (it can still use them, it just can't see them, so once you identify that machine you can isolate it and reroll the service credentials)
=> More informations about this toot | View the thread
Finally got around to setting up postgres "backups." (To a different disk on the same host. rsync to my NAS is TODO) Why does postgres streaming replication not support first-class bandwidth limits... I really don't want to simply throttle the fuck out of it directly with iptables... (I really wish docsis had better upload speeds...)
(Also, it's nice that pg_combinebackup exists, which allows for maintaining a rolling "base backup" for a retention policy, which is useful on small but high-turnover DBs, while still allowing for "every backup send is incremental")
=> More informations about this toot | View the thread
I may or may not have been nerd-sniped by lazy math in kyle hill's latest video and may or may not rabbit-hole into hypersonic drag curves at some point in the next 48 hours...
He was off by at least several orders of magnitude because he assumed:
=> More informations about this toot | View the thread
Stupid idea: Screw the treaties, I want a nuclear test Pascal B2. Not to test a new bomb design or anything, but just to see if that goddamn manhole cover actually made it to space.
Unfortunately we have insufficient data on the hypersonic characteristics of manhole-cover-shaped things, so napkin math is usually off by quite a bit more than a "fudge factor" due to that.
=> More informations about this toot | View the thread
While looking to improve my personal VN ripper tooling (which is entirely python, with some PIL-compatible C extensions for texture formats) I discovered that the leading public tooling for this particular game is currently written in PHP. Who the fuck writes a bytecode decompiler in PHP?
=> More informations about this toot | View the thread
I was just reminded of Linux's monumentally stupid design decision with swap and hibernation, that could be fixed really easily if someone wanted to do it: so you hibernate to a swap partition, right? Perfectly reasonable, since swap's structure is already well-suited for that. The stupidity is how it coexists with normal swap. The partition must be active, which makes sense, but there's no way to reserve space in or 'disable' an active swap partition's use as normal swap. So if you're running Firefox, the high memory load Firefox likes to maintain for cache reasons will quickly chew through swap space, and then you suddenly can't hibernate the machine when you close the lid until you close Firefox. Which also affects "hibernate on low battery" as well.
There is a reason that windows pre-allocates an estimated-size hibernation image, and it's completely separate from the page file...
=> More informations about this toot | View the thread
@eniko I managed to crash kitsune tails nice and hard on linux: The game apparently doesn't appreciate suddenly having no display outputs when I switch over my KVM, even for a second. Might have to hold off on playing it until they fix gnome in debian testing... (unsure if it's a game issue or an SDL2 issue, since the problem stack trace is in SDL2 pollevent)
(I'm currently running a bugged-out version of gnome that can just stop flipping surfaces for some reason, and quickly toggling the KVM gives it enough of a kick to come back, but also reliably segfaults the game. Yes, I know this is at least partly an xkcd 1172 issue due to me running a frankendeb...)
=> More informations about this toot | View the thread
So I finally figured out why my Chromecast couldn't talk to my jellyfin server (which is IPv6-only, like most of my lab) while my parents' Chromecast could reach it fine. Theirs is newer. Mine's a gen 1 or gen 2. The old ones straight up don't support v6. There are no MDNS broadcasts for it on IPv6 either, so it doesn't even have an address.
So it looks like I'll need to upgrade it, or just replace it with an actual computer... (It can't even do 1080p60 smoothly)
=> More informations about this toot | View the thread
From the other day that I forgot to post: I had heard that the US 7-11 brand is reworking itself to match the Japan one better, but I didn't realize that meant they'd start printing the price stickers in yen too... (In all actuality, it's probably just the machine being set to the wrong unit, given that this place is about as far away from Japan as you can physically get without ending up in the middle of the Atlantic)
=> More informations about this toot | View the thread
For context, every iPhone can use every transit card. US iPhones can use Suica et al. and JP iPhones can use SmarTrip cards without any problem, because apple hasn't added a completely arbitrary us-vs-them to presumably cheap-out on the licensing fees. I suspect that if google had wanted to support the feature (rather than what current evidence suggests, of "begrudgingly accept while complacently waiting for visa to crush everyone else under their monopoly") they could have negotiated "only pay the fee on devices that activate it" because it requires an external server to provision the card in the first place, so they know of every activation.
(also, a quick tip for the above-mentioned config bypass: FeliCa cards do not get erased by a factory reset, while mifare-based ones do, so you can get "the best of both worlds" by rooting, provisioning all the JP cards you want via osaifu-keitai directly, and then flashing back to stock and locking the bootloader)
=> More informations about this toot | View the thread
So I just got reminded about google being shitheads again: Specifically over the gpay HCE transit card functionality.
Specifically, they're gaslighting people in official support channels who are wondering why they can't add a Japanese transit card to a non-Japanese pixel, by claiming "The non-japan SKUs don't have the FeliCa chip." Except that statement is utter horseshit: there is no separate FeliCa chip. All that stuff is handled by the combo SE/NFC controller, which wouldn't you know, is exactly the same model across all SKUs and fully supports NFC-F and Mobile FeliCa. If anyone has an NFC-F reader (or a proxmark) and a Pixel, you can verify this by reading the IDm from it.
For absolute proof that the restriction is totally arbitrary, you can either modify a single line in a config file, or SKU-swap to a japan SKU, and wouldn't you know it, it fucking works just fine: https://github.com/kormax/osaifu-keitai-google-pixel
Also, bonus bullshit: they disable mifare-based cards like SmarTrip in gpay on japan SKUs.
=> More informations about this toot | View the thread
Finally picked up a copy of Celeste with the steam sale, and it's actually quite fun to play using a controller. My one criticism of the default mapping is that I'm used to "press and hold jump before you land" working as 'grab' for some reason, so it being mapped to left trigger is a bit weird to me. (also, is there a setting for "hold stamina only counts down while moving"? I get the reason for not having infinite climb, but sometimes I just need to reposition my thumb on the stick after a complex thing)
I had to switch from the duke to my 360 controller part-way through though, because I severely underestimated how bad the rubber was on my duke. my thumb just would not stay on the stick. That's not fault of the game's though, just that it's an OG duke and thus ancient by controller standards and showing it's age. Tempted to get a hyperkin duke if steam's runtime supports it, now that I know it fits my hands properly (the 360 controller feels tiny to me, and is only in slightly better stick-shape)
=> More informations about this toot | View the thread
So I just discovered something batshit about my phone... So I've been having intermittent signal drops while sitting at the same desk for hours, right? I finally figured out what it is: plugging in headphones with TTY mode enabled immediately murders the signal and then reconnects with EDGE at full signal strength. Turn TTY mode off (not sure how or why it got enabled) and suddenly full signal everywhere, even with headphones.
What the absolute hell is going on with TTY support in the telephony stack? Worst case you can literally do a slightly-modified bell 300 softmodem over the "voice" channel... Are they trying to do the T.38 fax thing with it, and it's not properly set up on LTE for most towers or some stupid shit like that?
=> More informations about this toot | View the thread
Random sysadmin gripe: we need a better way to manage "persistent machine credentials" rather than just "stuff a password somewhere in /etc and chmod 600". Like having a way for root to use samba's domain-bind "computer account" credentials to authenticate to things like log aggregation servers or shared NFSv3 mounts. (Specifically for authenticating "as" the computer, rather than sharing a service account password across all machines, or creating a new service account for all machines. For hosting things not tied to a given box, I'd still use the shared service account model)
=> More informations about this toot | View the thread
A potentially interesting fantasy world scenario I just came up with: a world in constant carrington-level aurora. On one hand, free wireless electricity with a long loop of wire, but on the other hand that electricity is both unpredictable and extremely temperamental. (Gee, sounds a lot like some magic systems...). The developments and restrictions on the tech in that world could be pretty interesting to play around with. And the timeline, too: would the trivial access to electricity spur an early development of the telegraph long before electricity was 'properly' understood?
=> More informations about this toot | View the thread
An interesting and mildly concerning problem I just realized: We effectively have no way to "from zero" bootstrap a new computer without another computer, unless we go digging through a computer history museum's collection for a machine with front-panel loading and/or a paper tape/punchcard reader. (Or an old parallel-ROM machine and a breadboard DIP-switch programmer)
Of course, it's extremely unlikely that would ever come to pass even in the case of a second carrington event, simply due to the number of 'old' machines unplugged and in storage, or ones deep underground on heavily isolated electrical systems, but it's still an interesting "what if?"
=> More informations about this toot | View the thread
So I just had a stupid idea. The NEC doesn't allow you to run low-voltage in the same conduit as high-voltage, right? Aside from one very specific situation: the entire low-voltage system is rated to withstand the highest voltage in the pipe (which includes being wired accordingly). That's how split-unit heat pumps can put their low voltage signals next to power: both ends are in "class 1" (high voltage) boxes, and the electronics are rated for it.
Well, what about ethernet? cat5e is generally insulated to 300v, which is fine, and the isolation transformers are rated for 1500v common-mode (if not using PoE) which is also fine. The problem? the connectors. They're generally in open-air and accessible to grabby meat-mittens. The solution comes in the form of PoE injectors: there's nothing saying that ethernet can't have more than two isolation transformers, which is how PoE injectors avoid back-feeding the device 'behind' them. 1/?
=> More informations about this toot | View the thread
So the IRS has apparently completely phased out their previous account system for ID.me. And the "upload a photo of an ID and take a selfie" requires a smartphone. And it's not that they're using an app, no, they're just using the standard browser capture APIs. Yet they won't just let me use the goddamn webcam I have strapped to my computer, which accomplishes the exact same thing. But it gets even dumber: if you click the "I don't have a smartphone" button, the only option is a video call using that same webcam they won't let me use to take a photo.
=> More informations about this toot | View the thread
I just saw someone on the train with a cat tree. An assembled cat tree. It's not that there's anything wrong with bringing a cat tree on a train, more just why?
=> More informations about this toot | View the thread
=> This profile with reblog | Go to becomethewaifu@tech.lgbt account This content has been proxied by September (ba2dc).Proxy Information
text/gemini