Lock @actions/cache to 3.0.6, later versions break bundler caching

[MSP-Greg]

See #483. I can't look at it until (hopefully) this weekend.

It may be a bug with @actions/cache, as I tried both 3.1.4 & 3.2.1. Not.sure.

[MSP-Greg]

Two jobs have 429 errors? That's 'Too Many Requests'? What else is going to happen today???

See:
https://www.iana.org/assignments/http-status-codes/http-status-codes.xhtml
https://www.rfc-editor.org/rfc/rfc6585.html#section-4

[MSP-Greg]

I did some more testing, and it seems that 3.1.0 and 3.1.1 also seem to work. 3.1.2 and later do not. Also, see:

actions/toolkit#1377

[simi]

Thanks for addressing this @MSP-Greg. Would it be enough to restart failing checks for now, 🙏 merge and release?

[eregon]

My thoughts is we should avoid "blindly" updating to latest versions of dependencies without a need, things tend to break, it's not the first time. Thank you for the PR.

[MSP-Greg]

My thoughts is we should avoid "blindly" updating to latest versions of dependencies

Given that the change in @actions/cache was a minor update (not a major update), it should work. And, CI passed.

I'll see if we can add some testing to caching to verify that it works correctly. That may be difficult, given that there may be CDN delays, not sure.

JFYI, this weekend I changed the code in computeBaseKey to setup-ruby-bundler-cache-v6 (from -v5). Using @actions/cache 3.2.1, it would save the cache, but never restore it.

[eregon]

This file has CRLF that's why the diff is so big, I'll regenerate it with LF.

[MSP-Greg]

Sorry & thanks. I thought I've got Git set up so that can't happen. Maybe a new version. I'll check. But, that file is CLI generated, and my editor is always 'LF' only.

[eregon]

I'll see if we can add some testing to caching to verify that it works correctly. That may be difficult, given that there may be CDN delays, not sure.

Maybe we can do that by using ruby/setup-ruby twice in in the same workflow job, and having some debug env var or so to fail if the cache isn't restored in the 2nd usage.

[MSP-Greg]

Maybe we can do that by using ruby/setup-ruby twice in in the same workflow job, and having some debug env var or so to fail if the cache isn't restored in the 2nd usage.

Yes, I thought something similar. I'm back to testing this because I'm just a little frustrated by it...