Fridtjof Siebert

The problem is not restricted to string constants, here is an example using a user defined feature `Obj`: ``` test_choice is Obj. first(A, B type, a A, b B) A...

Your examples only show errors at the end of a file, which is a more special case than end of a line. The question is, what is less confusing. In...

The behavious changed, now we get ``` > PRECONDITIONS=true POSTCONDITIONS=true ./build/bin/fz test_semigroup.fz error 1: java.lang.Error: require-condition1 failed: RefValue.java:74 "(!dfa._fuir.clazzIsRef(original._clazz), original._clazz == vc, dfa._fuir.clazzIsRef(rc));" at dev.flang.util.ANY.require(ANY.java:133) at dev.flang.fuir.analysis.dfa.RefValue.(RefValue.java:74) at dev.flang.fuir.analysis.dfa.Value.box(Value.java:347) at...

the added regression test shows that this is not fixed.

Updated example to current syntax: ``` a(T type) is b : a b is ``` this produces an error now (ambiguous type) before this hangs forever.

The example ``` a(T type) is b : a b is ``` no longer results in `fz` hanging, but it produces a confusing error that references the declaration of `b`...

Here, the type propagation should be improved to propagate the tuple type into the array elements. Given the type explicitly ``` ex is arr array (tuple (choice String (array u8))...

Still broken, here is a smaller example: ``` > PRECONDITIONS=true POSTCONDITIONS=true ./build//bin/fz -e "say unit.this.type" error 1: java.lang.Error: check-condition failed: Clazz.java:2181 "(Errors.any() || i == o || i != null...

The reason for this is probably different than that of #3391, so I re-open this.

I think you meant c : choice of a, b. I am not happy about the syntax using a dot here. Maybe just requiring a verbose a is b is...