ci: Remove CircleCI and update Node versions

[alexrudd2]

Closes #140.

I have forced an update of the transitive dependency nan in package-lock.json, but avoided listing it as a direct dependency in package.json. Along with updating webpack to v5, this would make the CI green (alexrudd2@725f0ed)...

... if #139 is fixed. So I expect this PR to fail for the time being.

[nornagon]

While removing CI does technically result in CI not failing, I'd prefer CI to get closer to passing instead 😅

[alexrudd2]

While removing CI does technically result in CI not failing, I'd prefer CI to get closer to passing instead 😅

Oh, sorry I was unclear. This removes CircleCI but retains/updates GitHub Actions. I'm assuming there was no need for both.

[nornagon]

Ah, right, yes that makes sense. Thanks!

I've regenerated package-lock.json because as a rule I don't merge package-lock.json changes from untrusted folks for security reasons.

[alexrudd2]

I've regenerated package-lock.json because as a rule I don't merge package-lock.json changes from untrusted folks for security reasons.

OK, fine with me. It actually might save a bit of trouble later in with broken dependencies.

Actually, I've now managed to update ALL the dependencies and fix the CI in my fork: https://github.com/alexrudd2/saxi. (It was a pretty painful introduction to TypeScript 😆 ).

How would you prefer I upstream things? (Note I did revert 593c451, as did other people).

[nornagon]

I don't want to accept a revert of 593c451 into this repository. I'm happy to review other reasonable changes though!

[alexrudd2]

I don't want to accept a revert of 593c451 into this repository. I'm happy to review other reasonable changes though!

OK. I'll gently ask you to reconsider. It does break the CI, apparently breaks some people's real hardware (#131), and at least one other fork has also reverted.

In any case, I'll see what I can do to apply the fixes I've learned using my fork in manageable PRs.

Are you interested in setting up renovatebot?

[nornagon]

The CI breaking is not a fundamental problem with WebSerial support; it's certainly fixable. It's far from clear to me that #131 is a problem with WebSerial per se. If it is, I'd rather find and fix the issue than remove WebSerial support.

I have no idea what renovatebot is or why I'd want it, perhaps you could give a little more info?

[alexrudd2]

The CI breaking is not a fundamental problem with WebSerial support; it's certainly fixable.

I was able to fix part of it (with a polyfill), but couldn't get it all the way. I'll send a PR with the partial fix.

It's far from clear to me that #131 is a problem with WebSerial per se. If it is, I'd rather find and fix the issue than remove WebSerial support.

I'll test on real hardware soon and let you know if I can reproduce.

I have no idea what renovatebot is or why I'd want it, perhaps you could give a little more info?

Renovatebot is a more powerful Dependabot. Here's an example PR - it groups similar updates together, and gives full changelogs: alexrudd2#42. I also set it up to automatically merge when the CI passes (with the revert, of course). No affiliation, I'm just impressed.

[nornagon]

Eh, no thanks, I like that dependabot is built into github and requires very little thought from me. Mend.io seems like something that quite possibly won't exist in 5 years, which is already less than the lifetime of this project :)

[alexrudd2]

Heh, good timing on dropping Node 12.x since GitHub Actions is going to drop it on Aug 14.