Notes for February 12-18

It was an interesting week, with a turn away from SBCs and a return to a bit of hardware-related coding.

Another late night meeting day, sporting Excel shenanigans.

• Re-printed the ball joints on my joystick-based spacemouse for a better fit. Realized there is still something wrong with the conversion matrix I’m using. Decided to try for a different approach regardless because the mechanical design is prone to wearing down quickly.
• Redesigned a few parts of another popular DIY spacemouse project to use an MPU-9250 as sensor and a WaveShare RP2040 Zero as controller, and did a test print with pretty good results:

• Decided to do a little unconventional experiment with Moonlight and got it working on the Quest 2.

Mardi Gras, so had the whole day to enjoy. I thus proceeded to goof off in the morning, family visits in the afternoon, hacking in the evening.

• Decided to go all out on virtual widescreen gaming:

• After wiring things and running a few I2C tests, my AliExpress-sourced MPU-9250 turned out to be an MPU-6050. I thus have no magnetometer (which should be fine, but was a conundrum for a while because I couldn’t find the AK8963 on the I2C bus).
• Realized I had designed the sensor mount mirrored, which warranted a quick re-print. The Two Trees SK1 is very addictive when you’re prototyping things.
• Spent a while navigating subtly broken CircuitPython libraries (why doesn’t an official Adafruit .mpy file released two days ago load on a month-old interpreter, both being stable 8.2.x releases?), ended up hacking my own IMU class out of other code on the Internet.
• Stopped short of implementing my own Kalman filter (for now). Will likely switch to C++ after a few more tests, after all…
• Grabbed this nice Gridfinity Parametric Generator and hacked together a customized version. Now my kids can order their own custom storage bins by just sending me a URL with the parameters they want.

Back to work, with a smattering of… Office work, of the Excel persuasion (forecasts are tricky). Once the day was over, changed things up and tried to go for more artistic pursuits.

• Received a Banana PI M7 (also known as the ArmSoM-Sige7), which seems pretty good. A brief test with this unofficial Ubuntu image was auspicious, although (as is becoming usual) it stopped booting after a mainstream kernel update. Ah well. I’m going to check out the official images over the next couple of weeks.
• Did some writing.
• Played around with NC Editor, trying to remember how to get the Novation Circuit Tracks to work in anything resembling “song mode”.
• Designed a couple of PETG brackets for a bathroom fixture.
• Played around with Reor, a note taking tool that can index a local Markdown repository and can leverage a local LLM. It sort of works with mixtral-instruct, but it currently mangles my files and does not support YAML front matter, so I’m passing for now.

Decided to revisit some coding projects.

• Spent a little while comparing .NET 8/Avalonia, SwiftUI and Godot as possible avenues for building a desktop tool. The lack of Godot documentation for building complex application layouts is a bit of a drag, while Avalonia, oddly, looks weird on a Mac.
• Another relevant datapoint: Most of the SwiftUI examples I can find are (as usual) outdated.
• Decided to stick with Swift for an hour during lunchtime to see how far I could get, was OK with the results but still far from a working application (let alone a cross-platform one, even inside Apple’s ecosystem).

I got a little test case working within the allotted time, but it took me a long while… far more than it ought to.

Sadly, I can’t get past the Mac Catalyst restrictions to actually set the wallpaper directly, so I might be upgrading this to an Xcode project.

Errand day. Actually went to the office as well.

• Did some more research and writing.
• Tried to get mixtral-instruct to flag and re-format some pieces of data, with mixed (ha!) results.
• Tried to replicate my Swift user interface prototype in Godot, without a lot of success–the layout model is just too different to get right on the first try.
• Toyed with lua-macos-app, which was just tantalizingly smooth (and a heck of easier to deal with given my Objective-C background and appreciation for Lua). That led me to Bounces, which seems amazing but, sadly, has a subscription licensing model.
• Fell asleep on the couch after another late night meeting.

Cleaning day, lazy idling.

• Spent a chunk of the morning messing about with my Quest 2 and Termux, of all things. And yes, you can run Visual Studio Code locally in a Quest device, without a web browser of any sort. I’m not sure why you’d want to though.
• Dusted the office and cleared out paperwork, spare parts, empty spools, etc.
• Spent a while testing and calibrating the Two Trees SK1 for a couple different filament types while I did some CAD work.
• Did a little research on various hardware manufacturers.
• Fiddled with accelerometer code again. Looked at several IMU handling implementations to see where I am getting my kinematics wrong.

Family day, had people over.

• Personal inbox zero, got back in touch with people returning from Spring Festival.
• Tidied office to make it less scary to outsiders.
• Started making a list of native development toolkits I like.
• Renewed one of my Azure Certifications (I know it’s Sunday, but I’d rather get it done now).
• Did some more SK1 testing and comparison prints.
• Tried wxLua, with considerable success.