Community
Chat Logs
Sunday, May 10, 2026
- bluephaleth: hi
- phalethhi blue
- blueso the point of my comment was that the current mechanism puts the colorscheme into a cookie. the reason for that is that on subsequent full page reloads, we read that cookie and can set the colour scheme in the HTML, so you don't get a weird flash
- blueif we remove the onHandle mechanism and the cookie, the mechanism is purely client side and you will likely see a flash
- blueI distinctly remember testing for this, so I'd be surprised if it suddenly worked without a cookie. but you'd need a full page reload
- phalethok, I will try that, but I think having the JS in the header is good enough to avoid flash
- blueJS in the header?
- phalethin the head tag is what I mean
- phalethscript tag in the html head tag
- blueah, right
- phalethhmm, you are right, there is a flash on palemoon
- phalethon epiphany it's flashing the same as current primate.run
- phalethI think then it's better to dump local storage and always use cookie
- phalethok, so it's already broken after one of my previous changes
- bluethe website is broken?
- phalethsort of :D
- phalethI got the fix
- phalethnow it does not flash even on epiphany
- phalethblue: how long should the cookie persist? 30 days?
- phalethI'll put in 9 days
- phalethso it's a single digit
- bluesure
- phalethok, didn't manage to trim much of the fat, but at least now there is no background flash at all on epiphany
- bluethat's great
- phalethjust little less than 0.5 kB of JS trimmed from the client bundle
- blueevery byte counts
- bluewe'll be doing the same for RP, anyway
- phalethyup, that's why I try to do all this, cause I noticed you started working on repopack website
- phalethanyway, thanks for the heads up on the flash
- blueyw
- phaleththe file naming is a bit incosistent
- bluewhere?
- phalethI've just fixed the naming and already pushed
- phalethso it's all good, btw I also got rid of extra requests to set the cookie
- blueok
- bluehow I test your PR, gh makes it so hard to switch to PR branches
- blueclanker is telling me to either install the gh cli tool (no go), or add your repository as remote
- phalethyou can hop onto the vps and deploy it there, I know that's like testing in production, but the changes are subtle
- blueyeah I wanted to test locally first
- phaleththen you can clone down my fork and checkout the branch
- bluemostly due to this line changing:
- blue <Icon name={$colorscheme === "dark" ? "sun" : "moon"} />
- blueof ot
- blueif it's not a store anymore, does it reliably change?
- phalethyes, it's just a state now
- phalethI've tested with librewolf, palemoon and epiphany
- blueok I'll clone your fork, but damn I hate gh for making this so hard
- bluewe gotta make this stuff easier with rp and item branches
- phalethyou can do it the git way, add a second origin and git fetch from that one
- phaleththen checkout the branch
- blueyeah I know, but this is dumb too. I don't want to add an origin for every contributor
- blueI seriously have no idea how this is supposed to scale
- phalethit's overcomplicated, but that's how it is
- bluewe can't accept that
- bluea golden path is a contributor having a branch like item/1 or item/website-optimisations and then I can just do git switch and test
- blueanyway, building the packages now
- blueso will test
- phalethcool, do you have brave browser or something else chromium based?
- blueyes, I'm on brave
- phalethnice
- blueman I hate pnpm
- blueprimate/apps/website on master via ๎ v26.1.0 pnpm install
- blueScope: all 55 workspace projects
- blueprimate/apps/website on master via ๎ v26.1.0 npx primate
- blueDone in 447ms using pnpm v11.0.9
- blueAlready up to date
- bluesh: line 1: primate: command not found
- blueseriously, pnpm is so bad
- blueand now I will delete the lock file and node_modules and run pnpm install and it will work
- phalethtry it from the root of the repo
- phaletheh
- blueya, already did, this is 100% pnpm being garbage
- phalethyou can do
- phalethpnpm purge
- phaleththat will get rid of all node_modules in the whole repo
- blueyes but this is a freshly cloned repo
- bluethis is really unacceptable
- blueok, now it's working after deleting everything and running pnpm install again
- phaleththey are going to fix it with rust
- blue100%
- phalethwill they succeed is the question
- bluelooks great
- bluepr is in
- phalethnice
- blueand we gotta move away from those PRs on RP. seriously
- bluethey're killing me
- dreamrealwhy?
- dreamrealwhat about them is killing you
- blueI spent around 5 minutes on setting it up, and 10 seconds on testing the code
- bluethat kind of overhead is 100% unacceptable
- dreamrealshoudl be a matter of git checkout, npm install (to update any dependencies), npm test
- dreamrealif that's not enough, someone's doing it wrong
- blueit's not, I'd need to add his origin to my git
- dreamrealthat's... part of the checkout, no?
- dreamrealI guess I should have been more specific (or less): "get the changes"
- phalethI think origin still needs to be added, just once for each contributor
- blueyes. and if I have 100 contributors, I need 100 origins, this is legit insane
- phalethkinda like upstream also needs to be added when working on fork
- blueall I wanna do is `git checkout item/website-improvements`
- dreamrealif only there was a CI/CD pipeline idea, you could run CI/CD and validate the code works without having to do a lot, then you could look at the tests and say "do they validate the change being represented? no? reject until they do, and until they pass"
- blueand that means for gh I either need ot use their `gh` command line tool, which I'm not willing to, or I need to add `git remote add phaleth git@github.com:phaleth/primate` for every contributor
- blueboth paths are dumb
- dreamrealat $work it offends the crap out of me to have my teammates say "CI/CD passed, let's test the code"
- bluebut I don't want to run CI/CD. I want to test the code locally. CI/CD is assumed to have already prerun
- dreamrealhas it?
- blueI don't have CI/CD, but that's not the point. the point is there should be an easy way for me to check out the contributor's branch, play around with code, look at it in the environment I'm best versed in, and push improvements upon it if necessary
- dreamrealyep
- dreamrealso identify what that means, and move towards it
- blueand I don't *want* that to be on some rando's repo which he could delete tomorrow, because that's weird, I want it to be as a branch on my repo
- phalethwe used to do CI/CD for primate, but it turned out to be too slow and sometimes there are build errors on master, which is alright, work is in progress
- bluegh totally destroyed the idea of forks. instead of properly using branches, every little thing is a fork of the whole repo
- phalethI think I can create a branch on primate repo, but only cause I'm collaborator
- bluethe funny thing is that issues and pr branches are already internally the same thing on gh
- blueso they're doing exactly what I'm talking about, but for some reason they're not exposing it to the surface
- blue(issue numbering and pr numbering are shared, that's how I know)
- phalethgitlab uses separate counters for those
- blueyeah, but github unifies them, which means it sees them as the same thing, which they are
- bluemy ideal workflow is easy. protect master from direct pushes, and have issues create server branches which assigned contributors can freely push to
- bluecollabrators can freely create issues and selfassign, externals can only take on approved work, or lobby for it first
- bluethat also prevents things like "why aren't you merging the PR I spent 2 weeks on without asking you if it was a good idea"
- blue1st rule of thumb: ask the maintainer if this is a good idea
- bluein this case, phaleth is a collaborator/maintainer so he'd have no problem creating and selfassigning an issue or just creating a dedicated longterm server branch
- blueanyway, back to RP hacking
- bluephaleth: this hot reloading is kinda lit, I changed the font in the static dir, and it immediately picked it out
- blueI can't tell you I know exactly why because I'm not importing the woff directly anyway, it is referenced in master.css but that isn't imported explicitly, I guess I have primate generate virtual imports for it
- bluepicked it up*
- phalethyeah, primate just picks up whatever is in static dir, does not care if it's used
- bluewe do have a js/css filter iirc
- blueso the fonts wouldn't be picked up automatically
- phalethyeah, I've added woff2 to that filter
- blueoh, that explains it then, look it me, pumping my chest
- phalethI think that's why font file names are now getting the content hash, kinda strange but works
- blueyeah
- phalethI break things by changing things
- phalethbut those are just side effects anyway
- blueya
- phalethblue: I'll remove the handlebars dependency from primate website to speed up builds, is that okay?
- bluephaleth: it's used for the rss, no?
- phalethyeah, I've managed to replace it with TS
- phalethI'll just open up a PR
- bluek
- phaleththe website redeployment method does not work, the read/write layer of the container gets wiped out after podman container restart
- bluedamn
- phalethmore reasons to speed up the build overall, tsgo was a good decision
- blueya
- phalethI'm thinking about removing svelte from primate website, google says there is over 22 kBs of unused JS (gziped), but it seems like there is more
- phalethsounds like a huge undertaking
- blueyou could probably replace it with htmx
- phalethyeah, or html
- bluehtml would be lit
- phalethheh