bug: awk parser doesn't treat newlines as statement separators between assignments

[chaliy]

Problem

The awk parser does not recognize newlines as statement terminators when the next statement is an assignment. Two consecutive assignments separated by a newline (no semicolons) cause an "invalid assignment target" error.

Reproduction

echo test > /tmp/t.txt
awk '
{
    x=1
    y=2
    print x, y
}
' /tmp/t.txt

Real awk: prints 1 2
Bashkit: awk: invalid assignment target

Adding semicolons fixes it:

awk '{ x=1; y=2; print x, y }' /tmp/t.txt  # works

Root cause

The awk parser likely treats the newline between x=1 and y=2 as implicit concatenation rather than a statement separator. It interprets:

x = 1
y = 2

as something like x = (1 y) = 2, which produces "invalid assignment target" because (1 y) is not a valid lvalue.

Per the POSIX awk spec, newlines are statement terminators equivalent to semicolons inside action blocks and function bodies.

Patterns that work vs fail

# FAIL: two assignments on separate lines
{ x=1
  y=2 }

# OK: semicolons
{ x=1; y=2 }

# OK: assignment followed by print  
{ x=1
  print x }

# OK: two if-statements
{ if (1) print "a"
  if (1) print "b" }

# FAIL: assignment after if (if last stmt is assignment-like)
{ if (1) x=1
  y=2 }

Impact

Critical. Nearly all multi-line awk programs use newline-separated assignments. This blocks the entire wedow/ticket script's awk processing.

[chaliy]

Investigation: The root cause is in parse_concat() and other expression parser functions in awk.rs. The skip_whitespace() function treats \n as whitespace, so x=1\ny=2 is parsed as a single expression with concatenation. Fixing just parse_concat isn't sufficient — the error propagates from parse_comparison and parse_assignment which also consume newlines.

A proper fix requires either:

1. Adding a skip_horizontal_whitespace() variant and using it throughout the expression parser chain
2. Preprocessing action block content to normalize newlines to semicolons before parsing

This is a substantial parser refactor — deferring to a dedicated PR.

---

Generated by Claude Code