Possible bug in sed(1) ?

D Spinellis zmact61 at doc.ic.ac.uk
Tue Feb 27 22:57:28 AEST 1990


In article <1567 at pttesac.UUCP> jlbrand at pttesac.UUCP (Jack Brand) writes:
>
>sed(1) gurus,
>
>I need help confirming whether I have found a bug in sed(1) or if
>I am misinterpreting the manual.

I presume it is time for a rerun.

From: diomidis at ecrcvax.UUCP (Diomidis Spinellis)
Newsgroups: comp.bugs.4bsd
Subject: Re: `t' command of sed does not work fine.
Message-ID: <760 at ecrcvax.UUCP>
Date: 23 Aug 89 10:44:54 GMT
References: <2427 at srava.sra.JUNET>
Reply-To: diomidis at ecrcvax.UUCP (Diomidis Spinellis)
Organization: ECRC, Munich 81, West Germany
Lines: 75

In article <2427 at srava.sra.JUNET> katsu at sra.JUNET (Katsuhiro Watanabe) writes:
>
>	Hello, world. I am using 4.3BSD on VAX.
> 	I found a bug on sed of BSDs, so I would like to report it.
>
>	Even if no substitution have been made before `t'(since
>the most recent reading of and input line), it often branches.
>
> [...]
>
>	Is this known yet?

I sent a bug report and fix to 4bsd-bugs at BERKELEY.EDU on May 18th 1989.
Here is the report and fix:

Subject: Sed does not clear the flag used by `t' when reading new input
Index: 	bin/sed 4.3BSD

Description:
	According to the sed manual page the test command `t'
	branches to the : command bearing the label specified
	if any substitutions have been made since the most recent
	reading of an input line or execution of a `t'.
	In the 4.3BSD version of sed the reading of an input line
	does not cancel the effect of substitutions made before
	reading the new line.  Thus if any of those substitutions
	succeded the first `t' to be executed will succeeed even if
	a new input line has been read.

Repeat-By:
	Execute the following:

sed -e '
s/hello/goodbye/
/goodbye/d
t n
a\
t failed
b
: n
a\
t succeeded' <<EOF
hello
world
EOF

	After reading the line containing `hello' the substitution will 
	succeed.  The d command starts a new cycle and thus the next t 
	should fail unless a substitution succeeds again.  After reading 
	`world' the substitution fails, but because of the bug, t succeeds 
	and the message `t succeeded' is printed.

Fix:
	Modify "bin/sed/sed1.c" as follows:

*** sed1.c.orig	Thu May 18 14:24:39 1989
--- sed1.c	Thu May 18 14:19:48 1989
***************
*** 650,655 ****
--- 650,656 ----
  {
  	register char	*p1, *p2;
  	register	c;
+ 	sflag = 0;
  	p1 = addr;
  	p2 = cbp;
  	for (;;) {

--
Diomidis Spinellis                  Internet:                 dds at cc.ic.ac.uk
Department of Computing             UUCP:                    ...!ukc!iccc!dds
Imperial College                    JANET:                    dds at uk.ac.ic.cc
London SW7 2BZ                      #include "/dev/tty"



More information about the Comp.unix.wizards mailing list