Mpv drops GNOME support (reverted later)

https://gitlab.gnome.org/GNOME/mutter/-/issues/217

> 4. Same as 1.

> ... saves tons of time

No. It really fucking doesn't.

How is it ever supposed to be faster to try and figure out which search result for 'index.html' is the file in the somewhat-cluttered folder I just fucking navigated to for the express purpose of opening index.html, so that I don't end up selecting another file named 'index.html' from one of the subfolders? Just jump to the goddamn file in the current folder as I type, it's so much faster, and it's what everyone who has ever used the type-while-in-a-file-dialog feature expects to happen anyway.

Also how have people not figured out that file search on your local computer kinda really sucks? The stuff you want to search for probably doesn't have a unique/meaningful file name. Folders namespace non-unique file names, and the full path then has some semantic meaning. And since humans are pretty good at semantics, we seem to be pretty good at navigating the semantic hierarchy of where we put stuff. Recursive search makes you read some inevitably tiny and low-contrast text to get the meaning back. Useful if you're truly lost but absolutely 100% not useful as the global default.

Also also, what the hell do we do with years (decades?) of muscle-memory? Pretty much everyone else still uses the old behavior too so if you're cross-platform (or use only some stuff that is GTK3) that muscle memory isn't going away.

Also also also, why the shitting fuck was #3 not the new default behavior (or why isn't the old default the new default for that matter). Why isn't this only-sometimes-useful recursive search bullshit hidden behind some non-intuitive undiscoverable hotkey?

All C standards since C99 have been ugly mishmashes of C++ features and bad ideas, one after another. What everyone was clearly asking them was to just copy-paste stuff from POSIX, and instead they came up with stupid nonsense like VLA or things nobody uses, like the [static N] syntax for arrays. I also still find absurd someone thought it was a good idea to add stuff as critical as threads.h without asking implementers first if they liked it (fun fact no one did).

What I think is a massive indicator of how much disconnected they are from reality is the "memory safe string handling" they added in C11 with those crappy *_s functions, which a. were just annoyingly slightly different from what POSIX and Win32 had and b. used a totally lunatic error reporting system based on callbacks and `set_constraint_handler_s()` - because obviously when a call to memset_s() fails I totally want a random function to run, for what for I still don't understand after 9 years. People just wanted the BSD's _l functions, and they gave them a memcpy() that called a random callback. No wonder C2X just wants to quietly pretend these abominations nobody ever implemented never happened.

What no one has ever had enough guts to admit is that 1. C strings, i.e. null-terminated strings, were a terrible mistake and 2. that C badly needs a new string library, or at least something that people can use without risking leaking their whole address space by accident.

It's more graceful than the program exiting in a bad way because you passively decided to stop support.