• I did the last break spec.
• constants specs all pass. There are tags, but it's only in 2.3 and we are not targeting that Ruby version yet.
• break needs a "frame on stack marker" technique to work most likely, like in SOM, so we can detect we should not throw the BreakException when the frame is no longer on the stack. And likely it's better to avoid this when we can detect statically it cannot happen.
• def might be a bit tricky, I'll take a look at it.
• The rest seems related to bugs/missing handling in the translator with masgn and other. Not very hard, just use bin/ast and breakpoint in the translator to see structure. setRHS and translateDummyAssignment might help to build assignments nodes with proper RHS.

constants specs all pass. There are tags, but it's only in 2.3 and we are not targeting that Ruby version yet.

If we can clear 2.3 tags that would be good. If we can get to zero tags then that's better than any other implementation isn't it?

Indeed, only MRI passes all of RubySpec currently.
I have some fix for that 2.3 constant change, but it's not trivial so I'd rather focus on 2.2 compat first (also 2.3 is still a bit of a moving target for some stuff).

The last character classes spec is blocked by jruby/jcodings#13

Last last regex modifiers spec is blocked by jruby/joni#13, but I've shimmed enough to pass the specs.

I can't see how to implement Global variable $0 actually sets the program name. This modifies argv[0], which we could do if only we could get the address of it.

I've figured out assignment of $0 on Mac. Still no idea for Linux. Note that prctl won't cut it as it's limited to 16 bytes.

Truncating for $0 might be OK, some OS only support in-place modifications of argv[0] I think.
Process.setproctitle also exists to set the process title (and it's also the name of a system call).
In MRI it's implemented in https://github.com/ruby/ruby/blob/trunk/missing/setproctitle.c for OS without setproctitle.

Yes, but the key thing we're missing is the value of argv. I can't find any way to do that on Linux.

For reading it?

Well, indirecting it in order to write to it, yeah.