it doesn't really work great for following squashes. not sure if i'm going to finish this or even submit it (at least not before a code review) but it's something to think about
let claude rip and it churned this out for me, a more convenient UI for doing interdiffs in a #jj-vcs repo. pretty simplistic but neat
he had some giant cry thread about it afterwords, but of course he couldn't give up on the classic insouciant churl Goon-persona while doing it so it didn't really go very well, as you can imagine
this reminds me that an old somethingawful guy still on twitter recently wrote a "joking" tweet that was suggestive he abused his ex-wife and then had a huge meltdown when people completely shat on him for sounding completely psycho lol
it is rather funny how much programmers argue about copyright given how little almost all of them know about it
need to get some traction on that pr of yours to configure the manifold urls so incremental upload works. i still use my janky home grown bullshit for my own repo but dislike having to run a watch daemon for upload
[after saying it ironically 400,000 times but this time actually meaning it] Is that good
I'm going to become The Joker dude, my god
if you don't use jujutsu then i immediately lose all respect for you 🫵 😹
It's out! jj 0.39.0! today! now! RIGHT NOW!!! github.com/jj-vcs/jj/re...
i've decided that today is a day to Eat Wings
mentioned this a few weeks ago but a genuine issue i don't see much talk about is the contradictions that lead to commodity fetishism of f/oss work. i assume nobody wants to do it because weird nerds will get mad when you suggest they are basically part of a fandom
Shaq Warns Young Players Not To Do Movie About Genie
Shaq Warns Young Players Not To Do Movie About Genie https://theonion.com/shaq-warns-young-players-not-to-do-movie-about-genie/
i forgot about this but in my monorepo like 7 months ago i just made it so every build retries 5x if it fails and it finally made auto merges reliable
issues like this probably have no impact on capacity planning or anything like that i bet
it's awesome that modern development environments are fragile enough that the default integration behavior for 99% of projects is just to pointlessly redo +90% of their work 1000x times a week
what the actual hell is going on with github actions recently man. it has gotten really bad
alas, it is github timeouts most of the time these days. recently the gap in e.g. p95 vs p99 times has been getting much worse it seems, so the normal amount of test time slack isn't cutting it
open source development is mostly about hitting the Retry Merge button because of flaky tests over and over again. that's about it there isn't really much else to it
not really just slightly different setup commands. it probably helps depending on your workload even on linux, but for jj on macos it’s the difference between completely unusable and fast
it's awesome how i can speed up tests on macos 1000x by just putting my tempdir in RAM so that it doesn't bottleneck on the filesystem. production operating system btw
now talking about the gold standard, silver prices, libertarian thoughts on money. fascinating stuff
Getting hit by a million lighting strikes after being depicted as a chud wojak by the evil voodoo device
Phones are the new voodoo doll, when someone makes you mad you just scream at it in the hopes that whatever God is inside punishes your enemy. maybe throw in a wojak for good measure, a kind of ceremonial art
it works pretty ok, but i'm not sure this is something i'm going to upstream without a bunch of thought... but i was able to go from a convo on discord about this to an actual implementation with ~15 minutes of spec writing, which is quite nice. closes the loop on problem space navigation a lot
the plan is badly written and half assed crap, but i wanted to see how it would turn out with at least a small bit of precision in the setup. the result is primarily 130 lines of (non-test) code and it works alright. i had to prompt for two bugfix follow ups: preserving metadata & a bug with merges
the tl;dr is you can do something like this:
jj git remote add foo-upstream github.com/bazquxx/foo
jj git fetch --remote foo-upstream
jj graft tree \
--from ::main@foo-upstream \
--path src/foo \
--onto ./vendor/foo \
--destination XYZ # default '@-'
and get a filtered history of "foo"
so yesterday i let Claude go ham on this, a `jj graft` command that can import subtrees of other repositories for some (limited) vendoring use cases. it was mostly able to one-shot this, followed by some small bugfix follow up prompts: gist.github.com/thoughtpolic...
he’s exactly what we need in these times
