← board

NilPy: s = "x" * 300 infers s as INT, then no overload of len matches

Found 2026-07-20 while landing [[bug-nilpy-string-local-truncates-at-255]]. Pre-existing — reproduces on the compiler built before that change.

Repro

s = "x" * 300
print(len(s))



Mismatch in MatchProcCall: name = len, nArgs = 1
  arg[0] = 1                      <- tyInteger
pascal26:2: error: no overload of len matches these arguments

print("x" * 300) alone is CORRECT, so the repeat itself lowers fine — only the inferred type of a local holding its result is wrong.

Cause (suspected)

The AST-based local typing takes ASTTk of the assignment's RHS, and the str * int binop node is tagged with the arithmetic result (tyInteger) rather than the string type — ParseTerm's tkStar branch types by the operand kinds and has no NilPy repeat case. str * int and int * str both need to yield the string type under PyExprMode, as does list * int.

Note

The diagnostic also dumps Mismatch in MatchProcCall: ... internals to stdout before the error line — that part is [[bug-overload-mismatch-dumps-internals-to-stdout]].

Gate

test-nilpy green + --tier quick + self-host byte-identical, with s = "x" * 300, s = 3 * "x", xs = [0] * 4 and their len() matching CPython.

Log