Mastodawn

Jun 14

Something's fishy. Two days in a row I hit semver incompatible #rustlang dependency updates. time-rs first, now dropbox's alloc-stdlib via brotli.

May 30

My take on the comman-and-question-mark topic: https://github.com/matze/shelley. Why? No huge ass bun or npm required for pi. And #rust. That's it.

P.S.: I also tried zerostack instead of pi but it was quite a hassle. I'll keep an eye on it though.

May 27

Tangled May 27

gm tanglers! its time for the release thread 🧵 ⭐ you can now access CI logs right from your terminal, with plain ol' SSH ⭐ 1. push your changes 2. if CI is triggered, the knot will give you an ssh command 3. paste that into a new shell 4. live tail CI logs! watch to know more: 1/5

I wrote about my async-first spinner and progress bar crate beginning of the year. I'd say it progressed far enough to give the ol' one-oh. Check it out at either https://github.com/matze/strides or https://tangled.org/matze.lol/strides (yeah, that's where my #tangled experience came from).

Max Leibman May 14

Every AI Bro: We are here to save humanity. We will start by making electricity, water, and electronics more expensive in order to build a machine to put you out of a job. You may begin showering us with gratitude now.

So far everything sounded pretty negative but I will continue test-driving it. I am not on board that GitHub departure train yet but on the other hand, I haven't been that excited (if you can say that) about a Git platform in a long while. And I have been running a Gogs instance since 2015 …

On to CI: #tangled uses "spindles", ephemeral container-based runners using nix as an environment. All in all, it works okay once you understand what nixpkgs you have to pull in. Most warts are again on the UX side: even though a bunch of pipelines have finished either successfully or with a failure, they are still displayed as "pending". Notifications about CI failures are not implemented yet.

On to pull requests and reviews. I am delighted that they do not follow GitHub 1:1 and having stacked PRs and jj usecases in mind. However, the workflow is a bit clunky at the moment. You push a new change (either modifying an old commit or adding one) and you must manually "re-submit" a new round for review. Reviewing the "rounds" is a bit confusing at first but makes a lot of sense. One weird wart about PRs: you just cannot change the title. It's probably some atproto thing.