Mastodon Feed: Posts

ireneista@irenes.space ("Irenes (many)") wrote:

alas, now we have to do that again because we didn't expect it to finish so we hadn't saved the output. we need to poke through it with a fine-toothed comb to continue from here :)

ireneista@irenes.space ("Irenes (many)") wrote:

yes! wow. it ran to completion. we're pretty sure there's some error messages mixed in there which make it not actually valid output, but this is way further than we expected to get tonight.

ireneista@irenes.space ("Irenes (many)") wrote:

we really need to replace these linked lists with hash tables, but that's one of those things we only really want to do after the bootstrap is done, you know?

the thing about having laid down every byte of this code by hand is we know exactly what performance bottleneck it's hitting as we gaze at the output lol

ireneista@irenes.space ("Irenes (many)") wrote:

wow. good news, it looks like it's gonna run all the way through the log-load transform now, as well.

in all the things we've done with Evocation to-date, this is the first time it's taken a noticeable amount of time to run anything to completion. so, we don't know for certain that it's working, lol, it just looks like it is...

ireneista@irenes.space ("Irenes (many)") wrote:

yep, buffer overrun. made it bigger ><

ireneista@irenes.space ("Irenes (many)") wrote:

huh. interesting. it starts to run, but a few hundred lookups in, it stops being able to find label values, even though it's looked up the same labels before and none have been added recently. probably memory corruption, based on prior experience...

soatok@furry.engineer ("Soatok Dreamseeker") wrote:

If you've encountered "flea" discourse about furry conventions already, you probably saw it elsewhere first and from a very bad-faith argument.

Here's something more grounded: https://soatok.blog/2026/09/08/multi-track-fursuiting/#fleas

ireneista@irenes.space ("Irenes (many)") wrote:

there's also a bunch of bugs where certain assembly instructions show their parameters in the wrong order, and stuff along those lines. that'll be a thing to go through carefully, but the immediate next step is to see if we can get it to also run on the log-load transform.

ireneista@irenes.space ("Irenes (many)") wrote:

wooooooo! it's able to run all the way to the end of the label transform, inside the hex transform, and produce output, without crashing

the output looks terrible, because it's almost entirely long blocks of addresses of Forth codewords that aren't annotated in any way because we haven't done that yet, but it exists

Boosted by soatok@furry.engineer ("Soatok Dreamseeker"):
catraxx@tech.lgbt ("catraxx :antifa:") wrote:

@baralheia It's stuff i've told people for a long time. Stop "cleansing" the fandom. We don't need to be "Good gays", we need to be ourselves. It is the only space on earth where we can do that. Stop trying to make us conform.

Boosted by soatok@furry.engineer ("Soatok Dreamseeker"):
catraxx@tech.lgbt ("catraxx :antifa:") wrote:

@baralheia also this https://youtu.be/UQAMmFzf-kM

Boosted by soatok@furry.engineer ("Soatok Dreamseeker"):
stevestreza@indieweb.social ("Steve Streza") wrote:

RE: https://mastodon.social/@nixCraft/117241250895962141

Ok, genuinely, tech journalists need to investigate why so much money is being laundered into Omarchy. A bunch of slopware from a white supremacist does not deserve $18 million when real Linux projects have gone unfunded for years. There has to be something else going on here.

Boosted by isagalaev ("Ivan Sagalaev :flag_wbw:"):
adampetrone@mastodon.gamedev.place wrote:

Remember kids, winners don't use GenAI.

Be a winner.

Boosted by neatnik@social.lol ("Neatnik :prami:"):
davidgerard@circumstances.run ("David Gerard") wrote:

"the robot will do your work for you" and "you can get rich for free" are the same promise and it always means they're going to pick your pocket.

ireneista@irenes.space ("Irenes (many)") wrote:

the good news is we're pretty sure we know what caused the crash we were working on before that

ireneista@irenes.space ("Irenes (many)") wrote:

bonus trick, if your compiler is named evoke and you usually do cat .......... | ./evoke with some extremely long incantation, and you do cat ..... > ./evoke instead, you get to test that everything still bootstraps cleanly lol

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

on the other hand i have no idea what is going on with the other thing going on here with jake and this like red violin situation or whatever is happening

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

the love story between odo and lwaxana troi is the sweetest love story ive seen in star trek yet. our sweet autistic boy gets married to intergalactic hyperempath diplomat dommy mommy and you couldnt ask for a better match
https://memory-alpha.fandom.com/wiki/The%5FMuse%5F%28episode%29

neatnik@social.lol ("Neatnik :prami:") wrote:

neatnik@social.lol ("Neatnik :prami:") wrote:

Two of the coolest Pokémon cards ever.

ireneista@irenes.space ("Irenes (many)") wrote:

less fun trick, at the shell prompt you can do gdb < a to redirect... you get it

ireneista@irenes.space ("Irenes (many)") wrote:

fun trick, at the (gdb) prompt you can do like run < a to redirect the subprocess's input from a file

ireneista@irenes.space ("Irenes (many)") wrote:

did you know that gdb has a mode where it opens up a little visual "window frame" inside the terminal?

we found this out by inadvertently sending a meg of source code to gdb as input instead of to gdb's subprocess. heh.

ireneista@irenes.space ("Irenes (many)") wrote:

sometimes when we're feeling particularly self-critical, we joke to our strategy is, we like to start with the empty program, and fix one bug at a time until it does what we want

ireneista@irenes.space ("Irenes (many)") wrote:

oh wow yes this is a much better strategy. it's showing us a partial and broken hex dump of some label-transformed code now. excellent.

cstanhope@social.coop ("Your weary 'net denizen") wrote:

I'm sure this flew by my timeline at some point in the past day or so, but I can't find the toot. I can't boost it, so I'll just tell you I enjoyed this talk from Felienne Hermans:

https://www.youtube.com/watch?v=0-6-f94n%5F9M

(I find it hard to squeeze in listening to a 48 minute talk, but at 2x, it's now a 24 minute talk, which is just barely manageable.)

Boosted by cstanhope@social.coop ("Your weary 'net denizen"):
ajroach42@retro.social ("Andrew (making toys)") wrote:

Tech is bleak, time to make some weird shit.

ireneista@irenes.space ("Irenes (many)") wrote:

okay, well, we actually decided on a minor change to the implementation strategy. we have this transformation-state structure and we wanted to turn it into a stack of itself, but managing the transitions between which layer was active when turned out to be quite brittle. we realized we can let the existing partitioning of allocation arenas do that for us, since code inside the hex transform has its own copy of most variables anyway.

Boosted by brib@bribstodon.xyz ("brib :neofox_floof:​ :Nonbinary:"):
serebit@floss.social ("Campbell Jones") wrote:

I won't explicitly say why Arch Linux refused DHH's sponsorship offer. It's not my place to do so.

But it shouldn't be hard to figure out.

Boosted by brib@bribstodon.xyz ("brib :neofox_floof:​ :Nonbinary:"):
thatgalsilver@zoner.work ("Syl (Sylvia Bun) :enby_inside::therian:") wrote:

As tech as a whole keeps getting worse, I'm investing more of my time into passion projects and older communities that are staunchly against a lot of the current tech movements.

Hate AI? Play more Doom. Hate mega corporations ruining everything? Play more Doom. The DoomWorld community is massively stuck in its ways, in a positive way. It's like jumping backwards 15-20 years in time.