fix: use custom HTTP server

[JustinBeckwith]

Not quite ready for primetime yet. Fixes #233.

[JustinBeckwith]

@XhmikosR this is handling everything I've tested it with, including bootstrap. Could I trouble you to give this a shot, and make sure everything looks right? It's a bit of a scary change 😆

[XhmikosR]

I checked out the branch, built it and replaced the local linkinator I had and it seems to work. Nice job!

# 2.11.2
C:\Users\xmr\Desktop\bootstrap>npm run docs-linkinator

> bootstrap@5.0.0-beta1 docs-linkinator C:\Users\xmr\Desktop\bootstrap
> linkinator _gh_pages --recurse --silent --skip "^(?!http://localhost)"

🏊‍♂️ crawling _gh_pages
🤖 Successfully scanned 705 links in 2.981 seconds.

# this branch
C:\Users\xmr\Desktop\bootstrap>npm run docs-linkinator

> bootstrap@5.0.0-beta1 docs-linkinator C:\Users\xmr\Desktop\bootstrap
> linkinator _gh_pages --recurse --silent --skip "^(?!http://localhost)"

🏊‍♂️ crawling _gh_pages
🤖 Successfully scanned 705 links in 3.006 seconds.

C:\Users\xmr\Desktop\bootstrap>

It feels like reinventing the wheel as long as it works fine :)

PS.

1. still getting some failures on my Windows VM like on CI
2. do you plan to switch to async/await for fs methods in server?
3. I would suggest adding a test with a folder with periods
4. should there be some kind of caching?

[JustinBeckwith]

Sweet. Looking at these:

still getting some failures on my Windows VM like on CI

Yeah I ran these on my windows machine, and the CLI tests just randomly will take 5 seconds to spawn a process. I have no idea what to do here, but I don't think this is new.

do you plan to switch to async/await for fs methods in server?

Done in the latest build

I would suggest adding a test with a folder with periods

lol good call, done

should there be some kind of caching?

So I thought about this, and no! The crawler caches all responses once it visits a url, so there's no way that you would need to load the same file twice. The cache would never actually get used - one of the many things in a normal static web server that I think it's easy to leave out here, because the only client for these requests will ever be gaxios/linkinator

[github-actions]

🎉 This PR is included in version 2.13.2 🎉

The release is available on:

• npm package (@latest dist-tag)
• GitHub release

Your semantic-release bot 📦🚀

[XhmikosR]

So I thought about this, and no! The crawler caches all responses once it visits a url, so there's no way that you would need to load the same file twice. The cache would never actually get used - one of the many things in a normal static web server that I think it's easy to leave out here, because the only client for these requests will ever be gaxios/linkinator

Yeah, I wasn't sure if it was needed either, just thought I'd mention it :)

One last thing, it seems this PR introduces a few further CodeQL issues. Not sure if all apply since I don't see all the details yet (when LGTM runs, I should be able to see them, but I don't have access to the security tab). One of them should be easy to fix by escaping any characters like you do elsewhere.