The value of the expression (which is itself determined within
scalar context) is first parsed, and if there weren't any errors,
executed in the lexical context of the current Perl program, so
that any variable settings or subroutine and format definitions
remain afterwards.