@crxsh.site.bsky.social
not a poster, i just play one on the internet irb(me):420> inspect - queer of gender, sexuality, mind (they/she or something like that) - π software/interfaces/language/radio - 240,000hrs of uptime - DM for signal or PGP/email info - ICE must be abolished
dozing on the couch watching a lets play of NFS3 ^..^
07.10.2025 20:57 β π 2 π 0 π¬ 1 π 0I wondered this, and thankfully ruby doesn't seem to have this exact problem (that would be egregious) and the problem is much more subtle
07.10.2025 05:45 β π 1 π 0 π¬ 1 π 0I think the reason this happens is float.exposed/0x418b0302d2... is exactly one float before 56647770.35 and an off by one error in the last place (ulp) is pretty common (not sure what IEEE754 has to say about this)
07.10.2025 05:42 β π 1 π 0 π¬ 1 π 0(sorry for orphaning this reply, I accidentally leaked my computer's username)
07.10.2025 05:29 β π 1 π 0 π¬ 0 π 0demonstration the 1.1 accumulation loop results in significant error at 56,647,770.3j
I was testing this empirically and... ??? that round(1) result is pretty suspect
07.10.2025 05:28 β π 1 π 0 π¬ 0 π 0every value is slightly overestimated when not printed exactly
ruby must be using a different rounding mode on the FPU to get these results
07.10.2025 05:18 β π 1 π 0 π¬ 1 π 0force with with a spelled out C-style for loop too:
i = 0
while i < yourboundhere:
# body
i += 1.1
(0r..20r).step(11/10r) do |i| puts '%f' % i end
oops I threw an accidental curveball and started the counter from 1, here's the exactly analogous:
07.10.2025 05:05 β π 1 π 0 π¬ 1 π 0oh wtf, there's no round(x, prec)??
I guess you have to resort to round(x*10)/10, ... and at that point your loop counter oughta just be premultiplied by 10
ruby code demoing (1r..20r).step(11/10r) do |i| puts '%f' % i end that produces exactly neat and proper summations of 1.1 repeating
but you don't gotta think about none of that if you have rationals in your lang:
07.10.2025 05:01 β π 1 π 0 π¬ 1 π 0if you math.round you'd get correct results (at least until you run out of significant figures, and idk precisely what the limit would be but you can analyze it with some algebra like
|0.1 - nearest_float64(0.1)|*n < 0.05
and solve for n)
(continuing from my other branch modestly defending js) yeah see exactly, the corecion rules of js is i think its biggest incurable problem and a terrible SE crime
07.10.2025 04:44 β π 1 π 0 π¬ 1 π 0there's worse SE crimes than this, Β±2**53 is enough integers for basically anything that wouldn't also risk overflowing 2**64 (though i do strive for excellence and wish every scripting lang had arbitrarily precision integers and rationals built in like ruby)
07.10.2025 04:43 β π 1 π 0 π¬ 1 π 0it would be so funny if this site had any signed 32bit overflow bugs please please
07.10.2025 04:32 β π 4 π 0 π¬ 1 π 0I appreciate all the people showing love and support to me. Iβm so grateful honestly. It means a lot. The ban isnβt that big of a deal for me. I think it was wrong and the explanation was bullshit. But thatβs ok and I accept. Social media comes and goes. Iβm grateful for all the connections I was able to make and all the people I get to be connected with off Bluesky. More importantly, the world needs more people to give a fuck. More people to give a fuck about Palestinians, more people to give a fuck about disabled folks, more people to give a fuck about our queer and trans siblings, more people to give a fuck about Black liberation. The world, and Bluesky, is full of bootlickers who will do anything to avoid doing anything to upset the status quo. These people are cowards groveling to those in power in hopes of protection. I just hope people continue to show love and care and kindness. Mutual aid is so important. Give when you can, boost when you can. We only survive this together.
Hear ye hear ye a message of hope and inspiration from our courageous leader in exile, Link. May all his loyal followers take heart!
06.10.2025 22:33 β π 2073 π 723 π¬ 20 π 34"you're not connected to anything, so no one can connect to you" (youtu.be/Y5BZkaWZAAA?...)
07.10.2025 04:12 β π 1 π 0 π¬ 1 π 0*EXTREMELY CORRECT BUZZER*
07.10.2025 04:08 β π 2 π 0 π¬ 0 π 0only way to be safe for sure
07.10.2025 04:06 β π 2 π 0 π¬ 0 π 1*snapping fingers*
07.10.2025 04:04 β π 1 π 0 π¬ 0 π 0bsky operators:
07.10.2025 04:03 β π 3 π 0 π¬ 1 π 0i would simply* prevent js from running within the iframe
*it would have to be preprocessed, there are libraries to do this because *email* is HTML sometimes (widely regarded to be a mistake that i'd be repeating)
i might have a poor impression of rss based off typical generator output but it could stand to be simpler maybe (an easy but nevertheless unconstructive critique of literally any well-adopted standard)
07.10.2025 03:42 β π 1 π 0 π¬ 0 π 0it is by far the coolest thing we've ever created and probably ever will
07.10.2025 03:40 β π 2 π 0 π¬ 1 π 0also can we like realize that the internet is its own world like there are literally no borders here and your governmentβs censorship is stupid and useless and harmful and has no benefits for the greater good
07.10.2025 03:37 β π 14 π 1 π¬ 1 π 0yeah, my first prototype would be a webapp that loads up <iframe>s on demand (making this secure is probably a nightmare though)
07.10.2025 03:37 β π 2 π 0 π¬ 1 π 0RSS can stick around for what its good at! the user experience i'm after would ideally be having a feed of arbitrary HTML documents from various sources, and i think maybe very little about RSS itself would need to change, and maybe this is more a project about a good RSS reader
idk im spitballin
Photo of a pixelated joint with the text youve been passed the virtual joint
06.10.2025 23:44 β π 48 π 12 π¬ 1 π 0I wish RSS wasn't so limiting, id kill for a redo of RSS
07.10.2025 03:24 β π 2 π 0 π¬ 1 π 0neocities is just webhosting yeah, it does have a veneer of social features though, you can follow other sites from your own to get site updates (an experience richer than RSS or atom, which is not hard)
07.10.2025 03:17 β π 1 π 0 π¬ 1 π 0