The brief told bees to reuse across modules, and that generates code which does not compile #4296

What I shipped yesterday, and why it is wrong

#4293 added a line to every issue the feeder writes: "If it exists, reuse it (use module::name;) rather than writing it again."

Cross-module reuse does not generate. Proven on a two-spec minimal case:

$ cat specs/shared_math.t27
module SharedMath { pub fn double_it(x: i32) -> i32 { var out : i32 = x * 2; return out; } ... }

$ cat specs/uses_shared.t27
module UsesShared { use shared_math::double_it;
    pub fn quadruple(x: i32) -> i32 { var once : i32 = double_it(x); ... } }

$ t27c spec-status specs/uses_shared.t27
IMPLEMENTED

$ t27c gen specs/uses_shared.t27 | grep -n '@import\|use double_it'
5:const std = @import("std");
18:// use double_it: no references in this module

$ zig test uses_shared.zig --test-no-exec
error: use of undeclared identifier 'double_it'

The same happens with the module-qualified form (use shared_math; + shared_math::double_it(x)): the qualifier is dropped and no import is emitted.

So the instruction asks a bee to write code that spec-status calls IMPLEMENTED and the oracle then refuses. That is the worst shape of advice: it passes the cheap gate and fails the real one.

Done when

• 1. python3 -c "import importlib.util;spec=importlib.util.spec_from_file_location('f','tools/queen/feed_empty_bodies.py');m=importlib.util.module_from_spec(spec);spec.loader.exec_module(m);print(m.PREAMBLE)" | grep -c 'do not try to import it' prints at least 1
• 2. the text states what a bee should do instead: name the file and line where the function already lives, and implement only what its own criteria ask
• 3. python3 -m py_compile tools/queen/feed_empty_bodies.py exits 0

Boundary

tools/queen/feed_empty_bodies.py