Squarespace used obviously AI-generated text to respond to a ticket of mine, using the sort of generic, patronizing platitudes that only an AI could muster, all to indicate "they understood how frustrating the situation must have been." No response would have been better.

I'm an ex-customer now.

Trivial!

I used github.com/cbrgm/bluesk...

as seen here: github.com/varabyte/kob...

You need to set two GitHub secrets and you're good to go.

Small release. Handful of fixes included, but released early because v0.23.1 broke anyone using the base path feature (kobweb.varabyte.com/docs/concept...). Sorry about that!

Interesting. I've DMed him before, it was a good interaction. When we find one of his replies for a question we are searching for, it's always a good answer.

But anyway, even more pressure for the Gradle technical writer -- make Vampire obsolete! I'm sure he wouldn't mind the break 🙂

There's a guy in the community, with the handle Vampire, who is so prevalent that I am almost certain you know who I'm talking about. He has done more to help us understand idiomatic Gradle than any other source. It should have been the official docs that hammered home this stuff though.

That's good to hear! I'll keep an eye out. I haven't reviewed the docs for a fair bit now.

I'm not really sure I can think of a specific request. I'm more imagining a full rewrite of the docs. A guide that is simultaneously informative and concrete in its examples.

And I'd love way better official Gradle docs and tightened up APId so writing such a blog post would be totally unnecessary 😛 (I actually appreciate they're at least working on the second part)

We now communicate between tasks when necessary by serializing into / deserializing out of a file (not sure why gradle doesn't support this as a first class concept). We hit snags with plugins that were not configuration cache friendly but that's better now. We updated to lazy APIs everywhere.

I was lucky in that I had a peer who did a lot of the migration. A lot of it was trashing my using `dependsOn` everywhere, pushing harder into chaining properties lazily with map and flatMap, and really understanding inputs and outputs.

Technically, we should avoid testing this stuff. That's Oracle's job. I might suggest injecting custom logic via the strategy pattern, and/or assert on some other aspect of the state.

Otherwise, maybe create a utility method that blocks until its own allocated weak reference is cleared.

github.com/JetBrains/in...

Great question. Someone like @gergely.pragmaticengineer.com might be in a better position to know, although this might be too small fish to attract much attention unless it's part of a bigger pattern.

"I just called the police a little bit, but since there's no risk, don't sweat it!"