http4s/http4s-jdk-http-client

Migration back to gh-pages

rossabaker opened this issue · 8 comments

rossabaker commented

The main site is down, and nobody else can see Netlify, so I migrated back to gh-pages so I'm not the single point of failure.

I've set up a CNAME, but GitHub still won't accept it:

> dig jdk-http-client.http4s.org +noall +answer -t A

; <<>> DiG 9.10.6 <<>> jdk-http-client.http4s.org +noall +answer -t A
;; global options: +cmd
jdk-http-client.http4s.org. 2894 IN	CNAME	http4s.github.io.
http4s.github.io.	2894	IN	A	185.199.108.153
http4s.github.io.	2894	IN	A	185.199.109.153
http4s.github.io.	2894	IN	A	185.199.110.153
http4s.github.io.	2894	IN	A	185.199.111.153

The /stable and /latest redirects will no longer work. I see a /latest directory in gh-pages.

If I ignore the security warning (which I expect will resolve itself), I get a 404. We might need an index.html or some other landing page that redirects to a particular version.

amesgen commented

I am working on a transition to Leika as part of #562, and Laika supports versioned documentation, so the need for redirects might go away as a side effect.

  • 👍1
rossabaker commented

Sorry to dump this on you abruptly, but everything's already broken, and I knew you were working on a refurb. I'll keep an eye on things, and if you need DNS help or don't have enough GitHub rights, please yell.

armanbilge commented

https://jdk-http-client.http4s.org/ is back :) still needs some follow-up, to link to the previous versions.

armanbilge commented

Things mostly looking up at https://jdk-http-client.http4s.org :)

The one big surprise was that the 1.x site encourages you to install 0.3.4! I think this is because you are using squash merges/cherry-picks rather than history-preserving merges.

  • 🎉1
amesgen commented

Hmm, how can this influence which git tag is considered to be the latest one?

armanbilge commented

It's because the history on main doesn't include any tags between 0.3.4 and 0.6.0-M1. (Ideally it should show the latest milestone instead of the latest stable, but this was a bug I fixed in typelevel/sbt-typelevel#198.)

For example, if we look at 9808298, which is the commit tagged v0.5.0 we see that it exists only on the series/0.5 and series/0.7 branches. It was "squashed" into main, so it lost the true history.

amesgen commented

Ah, that makes sense, I hadn't thought of that downside of just cherry-picking commits from main to the older branches like I did, properly merging everything would have prevented this 😭 . But with #586, more useful versions are now displayed either way, thanks!

  • ❤️1
amesgen commented

I think that the last remaining deficiency of the new site setup was fixed by #603 (the version in the sbt snippet on https://jdk-http-client.http4s.org/1.x is now correct), thanks @armanbilge!

  • 🎉1