I'm giving up on my live-updating postfix shell[1]...
# two-minute-week
k
I'm giving up on my live-updating postfix shell[1] for a year or three 😞 It's just too much work to create something that isn't a toy, that can handle loops with lots of iterations and so on. So I'm scaling back my ambitions to a more conventional REPL, albeit with one twist.. well, take a look. https://archive.org/details/akkartik-2min-2021-02-24 Amusingly, the fan kicks in halfway through, as my modern Macbook Pro struggles to run Qemu 🙂 Main project page: https://github.com/akkartik/mu [1] https://futureofcoding.slack.com/archives/C0120A3L30R/p1607329853147000
👏🏽 3
❤️ 1
😢 5
j
Kartik, no! The difficult work is the where the good stuff is hiding!
❤️ 2
💯 1
c
Video link isn't currently playing for me.
k
Interesting. I just tried it and I think I got one error where hitting play did nothing. Is that what you're seeing? Trying a second time seemed to work..
c
Nope; I get a spinning wait and then an error; tried a couple of times and just now again. Not something I've seen before.
k
So strange. Let me just host it on my website.
*

http://akkartik.name/pictures/akkartik-2min-2021-02-24.mov

(higher quality) * http://akkartik.name/pictures/akkartik-2min-2021-02-24.mp4
c
I got around to watching; thanks for the download version. It is nice that you can test because you are still in text mode. Those things are harder when you are trying to check graphics output.
💯 1
k
Yes. I started thinking yesterday about providing some of the OpenGL interfaces without any graphics acceleration. If you have any ideas for ways to make OpenGL testable I'm all ears. Is it even something worth having, or are tests not a good model for graphics code?
c
I think the best way to test GL is to have a database of expected images and compare your output with that. Then tolerate a small pixel % error. Update your golden images when you change the spec, etc. That's the only way I've seen it done.
It's tricky with GL, because different hardware draws things slightly differently too; maybe that's better these days, but it used to be quite a problem to validate the API when all the GPUs did slightly different rendering
k
@Jack Rusher The kids are not happy about my pivot. 3 year old after seeing the latest demo: I want the black numbers and the red cursor.
j
Reminds me of this article about Manu Chao recording his classic album, Clandestino:
'"Le hazard est mon ami" he likes to say – chance is his friend. They tried different different mixes on the children, aged three and six, who lived next door. "The ones they liked were the ones we chose," Manu recalled.'
🤔 1
b
"the fan kicks in halfway through, as my modern Macbook Pro"
No M1 @Kartik Agaram? Say it ain't so
k
That would save me for, what, 2 years? 🙂 I wouldn't possess a Mac if it was up to me. Especially now that Netflix works on Linux. The solution to shit software is not fancier hardware.
b
But you could argue that x86 microcode is shit software
And M1 replaces that with better software at the lowest level
I would love to be on straight nix but those apple people still do too good a job
though, i'm sure in a year or two something will come out that gives M1 a run for its money
but right now oh my goodness it's like swapping a taurus for a tesla
k
Fair! But still less open in toto. Also less repairable than my 9-year-old Thinkpad. I've been trying (and failing) to live by a new ideal that some things are worth putting up with inconvenience for. What we have today is not ideal. I'd love for RISC V to win. Or even ARM.
🍰 1
⚖️ 2
❤️ 1