Jeremy Bopp

The failure on macos with version 3.0.4 doesn't appear related to my changes. The windows failures on 3.1.2 may be up for debate. On macos and ubuntu, `IO#sysread(-1)` raise an...

The patch set has been rebased, and the specs that fail on Windows have been skipped for that platform. I included comments that point to the issues I opened in...

ruby/ruby#6354 will fix the sysread test that I disabled for Windows. I don't know yet if it will be backported. If it is not, what is the process to keep...

I think the fact that the schedule resource is deleted and then re-created is actually causing loss of backup resources that are tied to them when using the `useOwnerReferencesInBackup` setting...

ArgoCD, unlike helm, *will* perform CRD upgrades unconditionally. It uses `helm template` to render all the manifests in order to apply them itself, so it doesn't have helm's CRD upgrade...

Regardless of the method used to operate helm, the current hook will cause backups associated with a schedule to be deleted when the chart is re-applied for any reason if...

That would be great! Thanks.

We ultimately ended up doing the same thing in a wrapper chart, @lucagouty.

Yes. I requested that earlier in this conversation, and @jenting said that the feature could be added back in.

I believe this change effectively ensures that the following code is never executed: https://github.com/ruby/ruby/blob/eb7cc5b16c082e3535b1c4c2b8897eae030160da/io.c#L11266-L11286 If so, that code should probably be removed.