1#From dhw@gamgee.acad.emich.edu Sat Oct 31 22:54:07 1998 2#Return-Path: <dhw@gamgee.acad.emich.edu> 3#Received: from cssun.mathcs.emory.edu (cssun.mathcs.emory.edu [170.140.150.1]) 4# by amx.netvision.net.il (8.9.0.Beta5/8.8.6) with ESMTP id HAA08891 5# for <arobbins@netvision.net.il>; Sat, 31 Oct 1998 07:14:07 +0200 (IST) 6#Received: from mescaline.gnu.org (we-refuse-to-spy-on-our-users@mescaline.gnu.org [158.121.106.21]) by cssun.mathcs.emory.edu (8.7.5/8.6.9-940818.01cssun) with ESMTP id AAA14947 for <arnold@mathcs.emory.edu>; Sat, 31 Oct 1998 00:14:32 -0500 (EST) 7#Received: from gamgee.acad.emich.edu (gamgee.acad.emich.edu [164.76.102.76]) 8# by mescaline.gnu.org (8.9.1a/8.9.1) with SMTP id AAA20645 9# for <arnold@gnu.ai.mit.edu>; Sat, 31 Oct 1998 00:17:54 -0500 10#Received: by gamgee.acad.emich.edu (Smail3.1.29.1 #57) 11# id m0zZUKY-000IDSC; Sat, 31 Oct 98 00:16 CST 12#Message-Id: <m0zZUKY-000IDSC@gamgee.acad.emich.edu> 13#Date: Sat, 31 Oct 98 00:16 CST 14#From: dhw@gamgee.acad.emich.edu (David H. West) 15#To: bug-gnu-utils@gnu.org 16#Subject: gawk 3.0.3 bug report 17#Cc: arnold@gnu.org 18#X-UIDL: 7474b825cff989adf38f13883d84fdd7 19#Status: RO 20# 21#gawk version: 3.03 22#System used: Linux, kernel 2.0.28, libc 5.4.33, AMD K5PR133 (i586 clone) 23#Remark: There seems to be at least one bug shown by the demo below. 24# There may also be a Dark Corner involving the value of NR in an 25# END block, a topic on which the info file is silent. In gawk 26# 3.0.3, NR often seems to have the least-surprise value in an 27# END block, but sometimes it doesn't - see example below. 28#Problem descr: the log below shows a case where: 29# a) (this may be a red herring) the output of the gawk script 30# is different depending on whether its input file is named on 31# the command line or catted to stdin, without any use of the 32# legitimate means which could produce this effect. 33# b) NR is clearly getting clobbered; I have tried to simplify 34# the 19-line script "awkerr1" below, but seemingly unrelated 35# changes, like shortening constant strings which appear only in 36# print statements, or removing unexecuted or irrelevant code, 37# cause the clobbering to go away. Some previous (larger) 38# versions of this code would clobber NR also when reading from 39# stdin, but I thought you'd prefer a shorter example :-). 40#Reproduce-By: using the gawk script "awkerr1", the contents of 41# which appear in the transcript below as the output of the 42# command "cat awkerr1". Comments following # were added 43# to the transcript later as explanation. 44#---------------------------------------------- Script started on Fri 45#Oct 30 20:04:16 1998 chipmunk:/ram0# ls -l a1 awkerr1 -rw-r--r-- 1 46#root root 2 Oct 30 18:42 a1 -rwxr-xr-x 1 root root 47#389 Oct 30 19:54 awkerr1 chipmunk:/ram0# cat a1 #a1 contains 48#one printable char and a newline a chipmunk:/ram0# od -c xc a1 49#0000000 0a61 50# a \n 51#0000002 chipmunk:/ram0# cat a1 | awkerr1 #no surprises here 52#1 lines in 1 sec: 1 lines/sec; nlines=1 chipmunk:/ram0# awkerr1 a1 � 53#lines in 1 sec: 1 lines/sec; nlines=1 #?! first char is an uppercase 54#E-grave chipmunk:/ram0# awkerr1 a1 | od -N1 -xc 0000000 00c8 55# 310 \0 56#0000001 chipmunk:/ram0# cat awkerr1 #the apparent ^M's are not 57#actually in the file 58#!/usr/bin/awk -f 59function process(w) { 60 if(w in ws) { 61 printf " : found\n"; lc[p " " w]++; rc[w " " n]++; } 62 } 63BEGIN { 64 } 65/^/ {if(NR % 10 ==0)print "processing line " NR; 66 process($1); nlines++; 67 } 68END {p=w; w=n; n=""; 69 if(w)process(w); t=1; print NR " lines in " t " sec: " NR+0 " lines/sec; nlines=" nlines; 70 } 71#chipmunk:/ram0# exit Script done on Fri Oct 30 20:07:31 1998 72#--------------------------------------------- 73# 74#-David West dhw@gamgee.acad.emich.edu 75# 76