Fun fact I keep forgetting: Python's original `async`/`await` keyword implementation was written to use the same codepath as `yield from`. I wonder if that's still true...

So, is the TL;DR version of the PHP bugs "setjmp/longjmp on allocation failure tried to recover, but the accessible state from the longjmp handler is too f***ed to do anything reasonable but abort"?

640x480 should have been enough for everyone...

Pssst wanna buy some 4s?

https://www.youtube.com/watch?v=rfelvI_ikf4

(Uma Musume VAs can't get their visas for a concert in LA in 4 days. They applied 6 months ago. I've read- _only read_- that they had to secure the venue before applying for the type of visas they need it. So 6 months is literally the earliest time they could do it. Damn shame if the concert is cancelled.)

"Trump loses midterms b/c Honse Girl Concert was cancelled" is a wild sentence, but plausible now ._.

https://youtu.be/ffoQ0CPKh8Y?si=-jgNlNEUjVvbmKaS

I wonder why this works...

What do you mean I can't pickle OS resources? Why isn't that possible :P?

Not sure I can think of a single case where Python pickling has ever NOT gotten in my way... in any case, I have a proof-of-concept of unittest parallel that doesn't require a lot of code. Enough so that I'm willing to reimpl in other projects for now until I refine it.

Okay, if I fork this... I'm gonna need a name. It's gonna be a completely separate package tailored to my usage. Anyone got any catchy names for "minimal parallel unit testing"? Cause I'm stuck.

I don't feel good about the person I'm becoming by fighting and losing friends over AI. But embracing AI is a deeply personal attack over something I hold dear ("maybe we should remember how things are done before we had the luxury of not caring. And you can't put time/effort into how things are done, you should be cognizant of that").

There are some ppl who it is easy to cut people off or distance themselves. That's fine, and I wish I was one of them. But I'm not.

https://github.com/craigahobbs/unittest-parallel/blob/main/src/unittest_parallel/main.py Why would you use Claude for 400 lines of code... WHY?!

Thinking about forking from pre-Claude, but can't decide whether to write my own minimal version inspired by this or maintain a fork. I don't think anyone's gonna complain if I use a pre-slop version for now tho (I hope...).

I rewriting smolarith's docs from Markdown to ReST. No, this isn't actually a great use of my time. But I've had enough compat problems with Sphinx plugins and Markdown that I should prob get used to it.

(ReST isn't... bad... per say. It's just different and has loftier goals than Markdown alone. And muscle memory's a bitch.)

(This is putting aside that the old certification tests were deprecated, and transitioning to the new set of tests is... not simple due to mixed Python and Ruby usage. So I haven't done it yet.)

Anyways, Idk what to do here... probably will begrudgingly have Sentinel certified while bitching quite loudly about it. Like, I don't even want to work on Sentinel anymore, I'm so fucking pissed rn.

Real fucking glad we made it completely fucking impossible to have a new CPU arch succeed because our collective standards for codegen and programming languages are millions-of-person-years high.

https://github.com/riscv/riscv-arch-test/blob/act4/AGENTS.md

The RISC-V Certification Tests use Claude. Do I really say "I refuse to get Sentinel certified b/c the certification tests provided by RISC-V themselves are slop"? Christ Almighty...

(I'm also considering "Crimson", "Garnet", "Vermilion", and "Rose". Also, a darker variation of "Scarlet" works.)

I need a red shade as a Perfect Name for a piece of software... hmmm, Ruby works. Good thing there's no existing well-known software named Ruby...

Christ, I do NOT want to write another Amaranth simulator test wrapper... why couldn't pytest have just Not Gone To Slop?!