Out of 615 regression tests only 17 now fail to compile.  The majority of the failures are in large complex programs - mostly imp compilers (though pass1's seem to compile OK)

Anything using generic %name is not implemented yet.

Anything that requires a dopevector (mainly 2D array parameters) isn't implemented yet

There are a couple of actual code bugs that cause the compiler to crash, that I haven't started to track down yet.

The state of the new library, backtrace, signal, etc etc is anyone's guess.  I won't be putting much effort into those until after it compiles all the sources I can throw at it.

With all those caveats, I do think I'm close to saying you might try it on some simpler test programs.  Maybe not today but soon.  (I need to simplify how you would download and build it in a test subdirectory before I ask you to try it out.)

It's not going to be at the "let's throw tough examples at it as see if they work" stage, it'll be more like the "Wow, it does actually seem to run 'Hello World'" stage :-)