Incorrect setting of some '$' fields when FS = "" and LINT = 1

[Nick Hobson]

Hi,

I think I've found a gawk bug.  I have demonstrated it in gawk 3.1.7 under Arch 
Linux and in gawk 3.1.6 under Xubuntu 8.10.

The program is:

#! /usr/bin/gawk -f

BEGIN {LINT = 1; FS = ""}

{
    for (i = 1; i <= NF; i++) {
        a[$i]++
        print FNR, i, $i
    }
}

The input data file is:

abc xyz

When run as gawk-bug test-input > test-output, the expected output is:

1 1 a
1 2 b
1 3 c
1 4   
1 5 x
1 6 y
1 7 z

The actual output is:

1 1 a
1 2 \0
1 3 \0
1 4  
1 5 \0
1 6 y
1 7 \0

(where '\0' is a binary zero.)

So the problem is that $2, $3, $5 and $7 are incorrectly being set to binary 
zero.

Curiously, the bug goes away if:
(a) LINT = 1 is removed (or changed to LINT = 0),
(b) a[$i]++ is removed, or
(c) FS = "" is removed -- in which case the actual output is as expected:
1 1 abc
1 2 xyz

Attached: gawk-bug, test-input, test-output

Regards,
Nick Hobson

gawk-bug
Description: Binary data

test-input
Description: Binary data

test-output
Description: Binary data