One of my nodes kernel panicked in the middle of reinstalling linux-firmware to remediate Archlinux's "Manual intervention required" fuckup.

That was slightly terrifying.

Well that's a first in a long, long time. I drew a game I flat out could not finish because of emulator inaccuracy.

Blues Brothers 2000 on N64, enemies move way too fast on emulator, which is manageable if not a bit painful, until you get to a segment where you need to flee from an instant kill boss and, well.... yup.

I might be getting my villainex vibe drawn after all. Someone posted a pic they drew in a tiny IRC chat and the figure was like... exactly what I wanted and things happened.

Might be a tick before I get it though.

The universe conspires against vee in the weirdest ways...

https://www.twitch.tv/trysdyn/clip/ColdSwissPresidentTTours-yNyRBlbi9RpViakp

I have so much to say about the social landscape around all of this but I think I'll just leave it at "If Rhea Ripley wanted to, I would totally be down for her keeping that bodysuit she wore for her match holy shit"

I imagine that must suck to actually do physical performance in though. I think I recall Golddust saying as much at one point. The gimmick they forced on him and how much that costume sucked.

For vee it literally all hinges on "If the performer wanted to"

If I saw "Battle will commence in 24 seconds" I'd go "Are you serious?!" and corps-a-corps into the boss, I swear.

This is probably why I'm too pointy and "130 base speed" shaped for MMO raiding.

So the one issue that showed up in Licenseathon so far is interesting. Four times across 2 years now, I've hit a bug where a streamer has a network disruption sufficient to disconnect their OBS from origin and then it's able to immediately reconnect. Afterward, edges start pushing video that WebRTC players (browsers) can't handle and they start dropping exactly 20% of audio frames and will do that perpetually.

Digging into it, I think I found the root of it. The flow for how this works is RTMP comes in, its audio gets transcoded to Opus, this raw stream is pushed to the edges, where it's delivered as WebRTC. Aside from taking the AAC and re-encoding it to Opus (because RTMP demands AAC and WebRTC demands Opus; a step I want to eliminate one day soon), there should be no re-encoding.

What I wasn't aware of is the delivery protocol internal to my streaming server software seems to default to 44.1kHz, 128kbit audio streams so I was transcoding audio down unwittingly then transcoding it back up for delivery.

Turns out if you break an incoming connection for short enough time that the stream delivery to edges doesn't get torn down? One of those transcoding steps stops happening, and audio claiming to be 48kHz 160kbit, but is actually 44.1kHz 128kbit gets delivered to clients.

128 / 160 is 0.8. Exactly 20%.

That this played at all, even in its scrungled "0.2s out of every second is silent" too-fast squeaky state is shocking to vee.

Tapping the sign once again that says "The style of marathon where someone sits at an OBS and manually captures a Twitch webpage to restream your run is dying and with that comes a certain lack of flexibility in certain places so no, when you ignore the entire runner setup guide and stream to Twitch instead of the event RTMP server and go 'I dont get it cant you just capture this?', we can't"

It's a long sign.