@eregon There are some modifications to mspec's raise_error to support matching causes, modifications to existing specs to use that functionality, and some additional cause-checking specs here, if you want to review.

Failures appear to be GHA issues or bad timing in specs, unrelated to these changes.

Thank you for adding more cause specs.

My initial thought here is whether making raise_error more complicated for this is worth it:

-> { 1 / 0 }.should raise_error(ZeroDivisionError, cause:)
# vs
-> { 1 / 0 }.should raise_error(ZeroDivisionError) { |e| e.cause.should == cause }

It looks nice and it's expressive but OTOH it's not used in that many files.

The main concern is MSpec tries to be as simple as possible so it can be used by early Ruby implementations.
That means so far we don't use keyword arguments in MSpec (git grep -F '**' lib has none).
That's easily solved though, see https://github.com/ruby/mspec/blob/31e8f11c0bf18cf85fa3d3643ea8fc5a395fc3fe/lib/mspec/matchers/complain.rb#L4-L7 for example.

If you want to modify raise_error, could you make a PR https://github.com/ruby/mspec and add spec for it in https://github.com/ruby/mspec/blob/master/spec/matchers/raise_error_spec.rb? Or you could also do it in this PR and run the mspec specs locally.

It looks nice and it's expressive but OTOH it's not used in that many files.

I think it will be used in many more as we continue to specify language-level exceptions that should have causes. I've only added a few cases here.

And even without those additional cases, there were many places doing the block form of raise_error just to check causes. Those are almost all one-liners now.

@eregon I cleaned up the commit that adds the cause: functionality, added specs for it, and updated others to support the expanded error output. Also fixed all style issues.

All requested changes have been made, merging.