Warnings about unary minus might have been wrong.  It's a bit
surprising that nobody noticed yet ... probably the precedence-warning
feature hasn't really been used much in the field.

Rikard Falkeborn

Discussion: https://postgr.es/m/CADRDgG6fzA8A2oeygUw4=o7ywo4kvz26NxCSgpq22nMD73Bx4Q@mail.gmail.com