Module unreachable_code 

Unreachable code detection (PL406) Unreachable code detection (PL406)

Identifies statements that cannot execute because they follow an unconditional control-flow exit (return, die, exit, croak, confess, last, next, redo).

Algorithm

The lint uses recursive statement-slice analysis rather than a flat pre-order AST walk. This is the only correct approach: a pre-order visitor with a reachable: bool flag cannot distinguish “visiting a child of this node” from “visiting the next sibling”, so a return inside a nested subroutine body would incorrectly poison sibling statements in the outer scope.

The correct algorithm:

1. check_unreachable_code dispatches on the root, then calls visit_node for each statement in top-level lists.
2. check_statement_list iterates a &[Node] linearly. When an unconditional exit is found, all subsequent siblings get a PL406 diagnostic. Nested blocks are recursed into freshly.
3. Subroutine and method bodies (Subroutine, Method) trigger a fresh call to visit_node, so a return in an inner sub never affects the outer statement list.
4. eval { } blocks are intentionally not recursed into: die inside eval { } is caught and does not exit the outer scope.

Scope of detection

Unconditional exit	Detected?
return	Yes
die "msg"	Yes (direct FunctionCall at statement level)
exit $code	Yes
croak "msg"	Yes
Carp::croak "msg"	Yes
confess "msg"	Yes
Carp::confess "msg"	Yes
last in loop body	Yes
next in loop body	Yes
redo in loop body	Yes
return if $cond	No (conditional via StatementModifier)
die inside or	No (right operand of Binary, not a direct statement)
die inside eval { }	No (caught by eval)

Functions

check_unreachable_code

Entry point for unreachable code detection.