The problem is that syncing between devices is not implemented in KeePass itself but through an external tool (Nextcloud, Syncthing, or whatever else). The sync client will only see the ciphertext and won’t be able to tell which records have been changed, only that two different binary files have a common ancestor and are in conflict.

The most obvious solution is to lock and close the database when it’s not in use (which is a good practice from a security perspective too), and to sync immediately when it is changed.

ZSH still needs the completion data files to be installed. It won’t just magically know the completions.

Chemistry. Gets really dull without ions.

deleted by creator

tl;dr: yes, credentials are cached locally. https://github.com/dani-garcia/vaultwarden/discussions/4676

The major downside to the single file storage used by Keepass is that it’s easy to accidentally create a conflict between files on different devices if they’re not synced immediately. Conflicting files have to be merged manually or data might be lost. I’ve run into this several times with Keepass + Nextcloud. In comparison, a central master database with local cache can resolve conflicts between individual records.

deleted by creator

Megaman hasn’t even done anything wrong (yet)… but we HATES IT!

!clem

Half-Life Alyx 2

losing integration because “containerized”

Bollocks. I’ve seen that many times with Flatpak (can’t speak for Snap), and every single time it was either because the packager failed to set up permissions or because the user messed with permissions that the application needed. Break off the tip of a screwdriver and it will no longer function as a screwdriver.

And I know you’re talking out of your ass because AppImage isn’t even sandboxed.

taking GBs of space

That part is true and accurate, and for a very good reason: dependency pinning. System packages can break if they don’t have the correct versions of shared libraries. If a package requires a very old version of a library, and doesn’t link it statically or supply it with the package, it can misbehave, have missing features, or refuse to even start. Flatpak (and probably Snap too, can’t speak for it) solves that by letting the packager specify (pin) the exact version of a dependency. If five separate packages require five different versions of the GNOME application framework, then they will download five separate packages of the correct version. AppImage solves it by being monolithic: everything is packaged together into a single executable.

Tailscale should work. It uses Wireguard and does some UDP fuckery to get around the firewall and NAT (including CGNAT). I can stream Jellyfin through it at 1080p native with no significant buffering, it’ll work for music.

Look at this: https://github.com/ChthonVII/guildwarslinuxinstallguide?tab=readme-ov-file#part-11-toolbox

“Run EXE inside Wine prefix”

It might not work, though. I’ve tried to use a similar tool with Snowrunner, it found the process, but didn’t actually work.

One of these might help: https://steamcommunity.com/app/275850/discussions/0/601902145259725017/

Although I wouldn’t be surprised if it was caused by NMS’s own network backend. When the Corvette update was launched, multiplayer was fine on my work computer, but did not work on my home PC. It’s a constant dumpster fire.

If you have Steam installed, try putting them in ~/.local/share/Steam/compatilitytools.d.

Uh… kinda? Powershell has many POSIX aliases to cmdlets (equivalent to shell built-ins) of allegedly the same functionality. rmdir and rm are both aliases of Remove-Item, ls is Get-ChildItem, cd is Set-Location, cat is Get-Content, and so on.

Of particular note is curl. Windows supplies the real CURL executable (System32/curl.exe), but in a Powershell 5 session, which is still the default on Windows 11 25H2, the curl alias shadows it. curl is an alias of the Invoke-WebRequest cmdlet, which is functionally a headless front-end for Internet Explorer unless the -UseBasicParsing switch is specified. But since IE is dead, if -UseBasicParsing is not specified, the cmdlet will always throw an error. Fucking genius, Microsoft.

He can’t, he had to re-run a benchmark.

Never let perfection be the enemy of getting it to work.

Is this what normies feel like when Linux users tell them to just use Linux? I have some apologies to make.