Could you also cherry-pick the tests from #761 and adapt them so this PR also fixes #760? (I think it already does but it's not tested)

I'm +1 for this change but I still think we should disallow anonymous Data types from being dumped. It doesn't do what anyone would expect and will create a new Data class for each such object in the stream. I'll put together a PR.

FWIW, JRuby is not passing these tests. @nobu modified the build to pass even if JRuby is failing in #719. I'm not sure why.

The instance variable-related tests on JRuby will be fixed by jruby/jruby#9138.

Instance variables on Data subtypes is still pretty hacky and only supports immediate children, but the fix here passes all tests on this branch and instance variables on direct subtypes of a Data type work properly.

@eregon I cherry-picked your tests from #761. Your error messages were much nicer, but it's simpler to just use Data#initialize's exception. Also, I dropped the order-dependent test.

@headius I deleted that continue-on-error for this branch.

@nevans Thank you; I believe once my JRuby PR lands this should be green on jruby-head.

I restarted the failed jobs since I believe both JRuby and TR have pushed updated head builds.

Something very strange going on with the JRuby job on Windows but it's very unlikely to be related to this PR. I'll look into it tomorrow.

Nit: the comments in .github/workflows/test.yml should probably be updated.

Also if it's only JRuby on Windows failing I'd be inclined to exclude it for now, and reenable it when it's fixed (still better than no JRuby testing at all like on current master).

Re JRuby on Windows the failure is:
https://github.com/ruby/psych/actions/runs/20150642697/job/57857789416?pr=765#step:6:1925

   bad class file: D:\jruby-head\lib\jruby.jar(/org/jruby/Ruby.class)
    class file has wrong version 65.0, should be 61.0
    Please remove or make sure it appears in the correct subdirectory of the classpath.

Which means javac does not care about JAVA_HOME, only PATH (makes sense).
IIRC there was a discussion with @kou about this on some default gem, and I also found rake-compiler/rake-compiler#247 so maybe just a too old rake-compiler?

Actually rake-compiler is already 1.3.0 which has this PR:
https://github.com/ruby/psych/actions/runs/20150642697/job/57857789416?pr=765#step:5:32
I suspect JAVA_HOME: C:\hostedtoolcache\windows\Java_Temurin-Hotspot_jdk\21.0.9-10.0\x64 + File.join(java_home, "bin", "javac") might not do the right thing then, or somehow the File.exist? check fails.

I found the issue: rake-compiler/rake-compiler#247 (comment)
So I think we should exclude jruby on windows here and fix that separately, as it will likely need a fix in either jruby or rake-compiler.

I've rerun the jruby windows job after ruby/setup-ruby#836 so it should be green now, or at least proceed further than javac failing.

Nit: the comments in .github/workflows/test.yml should probably be updated.

I pushed a commit for that.

Thanks for the extra fix in setup-ruby!