← board

NilPy: unary minus binds LOOSER than // and % — -7 // 2 is -3, not -4

Found 2026-07-20 while landing [[bug-a-nilpy-floordiv-and-modulo-wrong-for-negatives]]. Separate root cause, left unfixed deliberately: that ticket was about the floor SEMANTICS, this is about PRECEDENCE.

Repro

print(-7 // 2)     # CPython -4  -> pxx -3
print(-7 % 2)      # CPython  1  -> pxx -1
print((-7) // 2)   # CPython -4  -> pxx -4     (correct with parens)

Silent, plausible, off by exactly one.

Root cause

The shared parser gives unary minus Pascal's precedence, where it binds looser than div/mod — -7 div 2 really is -(7 div 2) in FPC, so the Pascal behaviour is right and must not change. Python binds unary minus TIGHTER than // and %, so -7 // 2 is (-7) // 2.

Only NON-DISTRIBUTIVE operators expose it, which is why nothing caught it: -7 * 2, -7 + 2 and -7 / 2 give the same answer either way, so *, + and / all look fine. // and % are the whole observable surface today; ** will join them when it lands, and it will be worse (-2 ** 2 is -4 in Python).

Shape

This is the same category as [[bug-a-nilpy-and-or-in-unavailable-in-call-arguments]] and the rest of group 4: the shared parser's expression layers do not know NilPy's rules, because pyparser.inc owns only the bitwise and boolean layers. Fix the precedence where that group's design decision lands — a PyExprMode-gated unary-minus binding — rather than as a one-off, and cover ** at the same time so the answer does not have to be found twice.

Gate

make test + self-host byte-identical, test-nilpy green, and the three repros above matching CPython. Add a sweep of unary minus against //, %, *, /, + for both signs of both operands — the distributive cases are what hid this.

Log