[bug-gawk] Length and contents of RT may be wrong/garbled when RS==""

[Jeroen Schot]

Hello,

Below is a bug report from a Debian user. I have checked his findings
and the same behaviour exists in gawk 3.1.8 and 4.0.0. I have not
verified his patch (also attached). The original bug report can be
found at http://bugs.debian.org/619738

Regards,
-- 
Jeroen Schot

----- Forwarded message from Rogier -----
From: Rogier
To: Debian Bug Tracking System <address@hidden>
Subject: gawk: Length and contents of RT may be wrong/garbled when RS==""
Date: Sat, 26 Mar 2011 17:50:44 +0100

Package: gawk
Version: 1:3.1.7.dfsg-5
Severity: normal
Tags: patch

The contents of RT may be garbled and the length may be wrong when RS=="".

There are two cases:
- Case 1: The last record is 'terminated' with '\n' instead of '\n\n'
  In this case, the length of RT is reported as 0 instead of 1
  Example (1st and 3rd are OK):
    $ awk 'BEGIN {printf "0"; exit}' | awk 'BEGIN {RS=""}; {print length(RT)}'
    0
    $ awk 'BEGIN {printf "0\n"; exit}' | awk 'BEGIN {RS=""}; {print length(RT)}'
    0
    $ awk 'BEGIN {printf "0\n\n"; exit}' | awk 'BEGIN {RS=""}; {print 
length(RT)}'
    2
- Case 2: RT is longer than the shortest RT seen so far
  In this case, the additional characters in RT are garbage.
  In a non-C locale, the length is also reported incorrectly.
    $ awk 'BEGIN {printf "0\n\n\n1\n\n\n\n\n"; exit}' | LC_ALL=C awk 'BEGIN 
{RS=""}; {print length(RT),gensub("\n","\\\\n","g",RT)}' | cat -v
    3 \n\n\n
    5 address@hidden@
    $ awk 'BEGIN {printf "0\n\n\n1\n\n\n\n\n"; exit}' | LC_ALL=en_US.UTF-8 awk 
'BEGIN {RS=""}; {print length(RT),gensub("\n","\\\\n","g",RT)}' | cat -v
    3 \n\n\n
    3 address@hidden@
  In both cases, the output should be:
    3 \n\n\n
    5 \n\n\n\n\n

I have attached a patch that fixes these problems, and I have added some test 
cases
as well. The patched source passes all tests and compiles into a .deb without 
errors.
After applying the patch, execute permission must be set on the test scripts:
    $ chmod +x test/rtlen*.sh

I hereby put the patch, to which I have all rights, in the public domain, so 
that
there can (hopefully) be no legal objection to incorporating it.

Regards.

Rogier.
----- End forwarded message -----

gawk-3.1.7.dfsg.RT-patch
Description: Text document