Community
Chat Logs
Thursday, October 8, 2026
- blueWB nevet, jreicher.
- liolocblue: Hey blue! does the sqlite db package support a turso libsql database url?
- bluelioloc: hey; I'm unsure; the sqlite generally delegates to the native adapter
- bluethe sqlite package*
- blueI could check it for you
- bluelioloc: "No. @primate/sqlite currently accepts only a local database path or ":memory:". Its underlying @rcompat/sqlite client wraps node:sqlite or bun:sqlite, neither of which supports Turso鈥檚 libsql:// remote URLs."
- liolocah ok thanks
- blueIf you need this capacity, consider submitting a task to primate or rcompat. if libsql is standardised, this should be in rcompat, if not, in primate. can't promise implementation, but will consider
- blueif libsql:// is standardised*
- bluehttps://repopack.com/project/rcompat
- blueguest roles can report/read tasks
- liolocah ok, I will probably make my own module for this
- blueok, up to you
- liolocYeah I need it for a project i'm working on right now
- bluelooks like something @primate/sqlite could benefit from if it doesn't mean pulling in deps
- bluebbiab
- liolocinitially for speed I'll use an external dep, but if it works well i'll have an ai create it without dependencies
- bluek
- dreamrealhey, the signin feature works!
- dreamrealwoot!
- bluedreamreal: signin?
- dreamrealthe url you posted for repopack has a title of "sign in" which nevet did NOT echo because it failed the rule for "no signin pages"
- dreamrealfound an rcompat bug
- blue`task report --project rcompat`
- blue`task report` will *probably* be removed in the future and I'll just make it possible to use --project with all task commands, but for now it's still limited to reading tasks
- dreamrealOKAY, massive UI refresh coming up
- dreamrealyeesh
- dreamrealnot a lot visually changes, which is good
- dreamrealbut it's been all morning :D
- bluedid you report it, dreamreal?
- blueI don't see anything
- dreamrealShould be... primate task #183?
- blueI thought you said rcompat
- blueI'll look at it
- dreamrealoh, I think it's an rcompat bug, but ... found it in primate, so... err.... I AM A GENIUS
- dreamrealor something
- blueright
- dreamrealhttps://bytecode.news/atlas
- bluewow, awesome sauce
- blueyou konw, dreamreal, the one thing that takes me the longest is this review-revision cycle. I need to tell the agents to look at it. This is actually something that repopack chat solves: it will allow you to set a session to `autonomous`; so its agent (who is attached to a task) will be informed if attention is needed, either rereviewing, or revising
- dreamreal*nod* MAKE THAT NOTE and use it.
- dreamreal"Dangit, every time I run into ___ on $otherplatform it annoys me, I bet it annoys others too, I have something that doesn't annoy me"
- blueya well, it's just that they should talk to each other and work together. they don't need my input at these stages; if they did, they owuld stop
- dreamrealSure
- dreamrealjust saying that this is how you communicate the awesome sauce to others
- dreamrealI have the workaround for the problem but would far prefer it be in primate/rcompat, obv
- dreamrealBBIAF
- dreamrealthe atlas is what found the bug, FTR
- bluercompat and primate are both getting patches soon
- bluehttps://repopack.com/project/rcompat/tasks/4
- bluejust need to give this out of the door first
- bluewith omarchy.. could be good to have
- dreamrealno worries, and no hurry
- blue+1
- bluedreamreal: primate 0.42.2 is out
- blue+ rcompat patch, but that's not your direct concern, likely
- bluehi phaleth
- bluenub is such a dumpster fire, it accidently published a -pre package and marked it as latest...
- bluehttps://repopack.com/project/primate/tasks/185
- blueridiculous broken vibe software...
- dreamrealblue: doing a general tag cleanup now, which should end up touching the UIs, will update when it gets there
- blueand I couldn't even hit ctrl+c or ctrl+d on it. that's how bad it is, phaleth
- dreamrealright now tags are split between content and factoids, that needs to be normalized
- blueok dreamreal
- dreamrealand thank you!
- blueyw
- phalethhi
- blue"Nub asks for CLI OTP because it implements registry uploads itself. On EOTP, it invokes a terminal prompt; it has no browser authentication flow and does not retain the prompted OTP across recursive packages. Pnpm uses its web-auth implementation, opens the browser, and polls npm鈥檚 completion URL."
- phalethblue: oh, really, so pnpm is still needed for publishing
- blueit's so bad, having to enter >20 times the OTP one after another
- blueoh man I was so happy to get rid of pnpm
- dreamrealtag normalization is a necessary thing in nevet, factoids and content have different taxonomies and they shouldn't :/ resolving them is dumb
- blueand now nub is just a dumpster fire
- phalethwell, if it's just publishing for which you need pnpm then I guess that's still not too bad outcome
- phalethI think I saw some issues regarding publishing on the nub issue tracker, but it seems like the author does not focus on that currently
- bluewe need flog
- bluelike, yesterday
- blueeverything sucks
- blueit's like, every time you want to mvoe a step forward, you move two steps back
- blueso for 0.42, I can still publish with pnpm
- bluebut we do need a solution for master
- blueI'll see if we can keep pnpm for publishing, but two package managers is annoyig
- phaleththere is also still bunx in a bunch of places
- bluein primate?
- phalethanyway, if you just put pnpm in package.json script, you can call it with nub
- phalethyeah, in primate
- phalethor npm even if you can't stop getting used to typing nub for some reason
- phalethnpm run publish
- bluehttps://repopack.com/project/primate/tasks/185
- bluethis smells
- blueno, npm can't publish workspaces, phaleth
- bluenot to my knowledge, anyway
- bluehive mind: "So Nub builds; pnpm packages and publishes. The browser authentication and registry interaction are handled by pnpm."
- phalethI mean npm run publish runs the publish script that executes pnpm
- phalethbut can as well type nub run publish to execute pnpm
- blueya
- phalethok, I see that clanker already wrote some long TS script there
- blueyeah there's going nowhere
- bluewe're either going back to pnpm+ts check until nub gets serious, or clanker will find a way
- bluebest solution is, we get rid of both, they're both bad
- phalethnot sure it's worth adding all that code, can just modify the script line in package.json to use pnpm
- phalethoh, so you got rid of pnpm-workspace.yml
- blueyes
- blueI moved to nub completely, with nub.lock
- bluenow I'm regretting it
- phalethyeah, I bet, I didn't want to do that
- dreamrealAh, the joys of committed development :D
- phalethI wasn't expecting a gotcha, but it's always better to be cautious
- blueI'm just sad everything is bad
- bluethis is why I end up writing my own stuff. not because I think it's good, but because other tools just suck
- bluethey don't do the basic things you expect them to do
- phalethwell, I recommend reverting primate#178 and in primate#185 package.json modify "publish": "nub publish -r --access=public", to publish": "pnpm -r --access=public publish",
- phalethand also in primate#185 get rid of all that clankered nonsense
- phalethalso no reason to be sad about software tools, it's just tools not always perfect
- blueI think I'll revert nub completely if pnpm can run TS checks
- blueand then we just move to flog when it's ready
- blueclanker is on it
- blueit's checking why pnpm didn't do TS checks earlier
- phaleththere was no flag like --noCheck or the update --no-check I could find
- phalethbut yeah, you may as well go back to pnpm fully, which means slower clankering I think
- phalethupdated*
- blueyeah, checking
- phalethdeployment of primate website does not have to be done using pnpm, so using pnpm 12 is ok
- blueI think the next codex limit will be dedicated to getting flog to do package manager stuff
- bluebecause the landscape is awful
- phalethwait, look, it's rustified already https://github.com/pnpm/pnpm
- nevetGitHub - pnpm/pnpm: Fast, disk space efficient package manager
- blueoh god no god save us
- bluedreamreal: we need help
- phalethheh
- bluestarted work on flog
- bluephaleth: we probably need to prepare flog.run soon :P
- phalethyou are really gonna create a flog project?
- bluealready did
- phalethI think repopack and also primate still need a lot of work
- phalethand you gonna use the zig rewrite from github?
- blueya, everything needs a lot of work, but blockers should be unblocked. I'm giving this one day
- blueno, I'm gonna go with go. I kinda lost faith in the zig project
- phalethok, there is a js runtime written in go that's kinda popular
- bluethere is?
- blue(note I'm limiting flog for now for pure package-manager stuff, not runtime)
- phalethanyway, I'd recommend giving primate more effort, primate#186 makes sense
- phalethah, ok, that's good
- phalethbut you can just finish primate#186 and move on
- blueyeah. this is a spike, phaleth. if it takes longer than a day or two, I'll give it up and go back to primate#186 or its follow-up
- blueI don't intend to spend a month on it
- blueI just wanna untangle this situation if I can. two package managers is annoying, undoing the nub work is annoying, pnpm is annoying too
- blueso let's see if we can find the golden path
- blueI have a few other nitpicks I've had with package managers along the years, would be a chance to fix them too
- dreamrealthe answer is to use rust like god intended! ... or go! ... or python! ... or whatever today's favorite language is
- dreamrealI can't keep up
- phalethwell, pnpm 12 is prolly not very slow after all
- bluephaleth: sent you an invite to flog project
- phalethok, thanks
- bluehttps://repopack.com/project/flog/tasks?status=open
- phaleththe runtime is called goja, here's some performance comparison table https://github.com/dop251/goja/issues/2#issuecomment-2763265666
- nevetBenchmarks and performance data? 路 Issue #2 路 dop251/goja
- bluenot bad, phaleth
- phalethyeah, but the runtime needs perf, look at where ladybird sits in the table
- phaleththey prolly managed to optimize something during the year since the table was put together but still
- bluewhat's the point of putting bun in that table?
- phalethactually 1.5 years already
- blueheh, deno is slower than node
- bluewhat a stupid table, sorry
- blueconflating runtimes and engines
- phalethheh, yeah, but I think that's ok
- phalethshows engines are faster, obviously, but also by how much
- blueI wonder why should a runtime add any overhead at all
- bluewell, any meningful overhead, anyway
- phalethit's cause of FFI
- phalethv8 is C++, but deno is Rust
- phalethand node is C++, but they prolly also have some bindings
- phalethjust so they can easily update v8
- blueso you're saying, always write your engine in the same language as the runtime?
- phalethno, C++ is a mess
- blueyessir
- phalethalso don't write your engine, too many problems with web standards
- bluedreamreal: do we need a js engine/runtime in java?
- dreamrealWe already have one, depending on what you mean by "in java"
- dreamrealwe have multiple, actually
- phalethjava has nashorn
- blueone that can compete with v8
- dreamrealand graal has a different one, and there's rhino, too, if memory serves
- phalethnashorn cannot compete with v8
- phalethyeah, rhino is like a predecessor to nashorn I think
- dreamrealgraal MIGHT be able to but I doubt it, using JS from within java isn't really "the way to do it"
- blueheh
- dreamrealOKAY! Now logs on pudl can be filtered to be ONLY conversation to cut out all the join/part/quit stuff
- dreamrealAnd now primate has the log filter in place too!
- dreamrealAight, I'm off for the day, seeya
- bluebyebye dreamreal, have a good one