Broken type inference - (list ?? anotherList) should be inferred as iterable

[natebosch]

(list ?? anotherList) is inferred as EfficientLength which is not iterable so the following snippet results in an analyzer warning:

for(var _ in (list ?? anotherList));

as does

var newList = list ?? anotherList;
for(var _ in newList);

but the analyzer can be appeased by

List newList = list ?? anotherList;
for(var _ in newList);

Ideally all 3 of these cases should work with no warnings.

[bwilkerson]

Yep. That's the result of the way least-upper-bound is defined in the specification. Strong mode has a slightly different definition that tends to do what you want.

[natebosch]

For the record: this behavior exists running in strong mode as well.

[jmesserly]

Yup, we want to fix this in strong mode! Thanks for the report.
Merging this into #25821