From owner-freebsd-bugs@FreeBSD.ORG Thu Dec 11 09:00:07 2008 Return-Path: Delivered-To: freebsd-bugs@hub.freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id C70C41065672 for ; Thu, 11 Dec 2008 09:00:07 +0000 (UTC) (envelope-from gnats@FreeBSD.org) Received: from freefall.freebsd.org (freefall.freebsd.org [IPv6:2001:4f8:fff6::28]) by mx1.freebsd.org (Postfix) with ESMTP id B75FD8FC18 for ; Thu, 11 Dec 2008 09:00:07 +0000 (UTC) (envelope-from gnats@FreeBSD.org) Received: from freefall.freebsd.org (gnats@localhost [127.0.0.1]) by freefall.freebsd.org (8.14.3/8.14.3) with ESMTP id mBB9078Z084415 for ; Thu, 11 Dec 2008 09:00:07 GMT (envelope-from gnats@freefall.freebsd.org) Received: (from gnats@localhost) by freefall.freebsd.org (8.14.3/8.14.3/Submit) id mBB907WZ084414; Thu, 11 Dec 2008 09:00:07 GMT (envelope-from gnats) Date: Thu, 11 Dec 2008 09:00:07 GMT Message-Id: <200812110900.mBB907WZ084414@freefall.freebsd.org> To: freebsd-bugs@FreeBSD.org From: "Garrett Cooper" Cc: Subject: Re: bin/129566: behavioral change of "read" builtin for sh on 8-CURRENT X-BeenThere: freebsd-bugs@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list Reply-To: Garrett Cooper List-Id: Bug reports List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 11 Dec 2008 09:00:07 -0000 The following reply was made to PR bin/129566; it has been noted by GNATS. From: "Garrett Cooper" To: "Michael Proto" Cc: FreeBSD-gnats-submit@freebsd.org Subject: Re: bin/129566: behavioral change of "read" builtin for sh on 8-CURRENT Date: Thu, 11 Dec 2008 00:52:33 -0800 On Wed, Dec 10, 2008 at 10:57 PM, Garrett Cooper wrote: > On Wed, Dec 10, 2008 at 8:07 PM, Michael Proto wrote: >> >>>Number: 129566 >>>Category: bin >>>Synopsis: behavioral change of "read" builtin for sh on 8-CURRENT >>>Confidential: no >>>Severity: serious >>>Priority: medium >>>Responsible: freebsd-bugs >>>State: open >>>Quarter: >>>Keywords: >>>Date-Required: >>>Class: sw-bug >>>Submitter-Id: current-users >>>Arrival-Date: Thu Dec 11 04:40:01 UTC 2008 >>>Closed-Date: >>>Last-Modified: >>>Originator: Charlie & >>>Release: FreeBSD 8.0-CURRENT i386 >>>Organization: >>>Environment: >> System: FreeBSD minibsd8-dev.localnet 8.0-CURRENT FreeBSD 8.0-CURRENT #0: Thu Dec 4 15:58:57 EST 2008 root@minibsd8-dev.localnet:/usr/obj/usr/src/sys/MINIBSD8-DEV i386 >> >> >> >>>Description: >> I've noticed a behavioral difference of the "read" builtin statement within /bin/sh on CURRENT and I'm hoping someone can point me in the right direction on how to restore the old behavior. >> >> I have a /bin/sh script that accepts input for IP address information: >> >> #!/bin/sh >> set -x >> DEFINT=vr0 >> DEFIP=192.168.0.1 >> DEFMASK=255.255.255.0 >> read -p "Enter network interface [$DEFINT]: " -t 5 INT >> read -p "Enter IP address [$DEFIP]: " -t 5 IP >> read -p "Enter netmask [$DEFMASK]: " -t 5 MASK >> echo ${INT:=$DEFINT} : ${IP:=$DEFIP}/${MASK:=$DEFMASK} >> >> >> This script waits for terminal input for each of the above variables (INT, IP, MASK) and defaults to a script-provided value if no input is entered in 5 seconds for each variable. On 6.x and 7.x if I simply hit Enter at the prompt (and don't provide any input) no value is assigned to the variable so my INT, IP, and MASK variables are set to null and the parameter substitution of the DEF* variables works as expected. >> >> On 8-CURRENT if I hit Enter with no input at the prompt the system seems to recognize the newline as input and continues to sit there until I hit Enter again. When I do this there appears to be a strange unprintable value assigned to the INT, IP, and MASK variables yet the variable subsitution doesn't work correctly. >> >> The man page on sh lists the following behavior for read: >> >> read [-p prompt] [-t timeout] [-er] variable ... >> The prompt is printed if the -p option is specified and the stan- >> dard input is a terminal. Then a line is read from the standard >> input. The trailing newline is deleted from the line and the >> line is split as described in the section on White Space >> Splitting (Field Splitting) above, and the pieces are assigned to >> the variables in order. If there are more pieces than variables, >> the remaining pieces (along with the characters in IFS that sepa- >> rated them) are assigned to the last variable. If there are more >> variables than pieces, the remaining variables are assigned the >> null string. >> >> >> As I interpret this, read should delete the trailing newline and assign a null value since there is is no "input" before the newline. Then the variable substitution should take over and assign the DEF* variables appropriately. 6 and 7 follow this but 8 does not. >> >> Here's the output of the script (with set -x) to help show what I'm seeing. >> >> This is on 6 and 7: >> >> + DEFINT=vr0 >> + DEFIP=192.168.0.1 >> + DEFMASK=255.255.255.0 >> + read -p Enter network interface [vr0]: -t 5 INT >> Enter network interface [vr0]: >> + read -p Enter IP address [192.168.0.1]: -t 5 IP >> Enter IP address [192.168.0.1]: >> + read -p Enter netmask [255.255.255.0]: -t 5 MASK >> Enter netmask [255.255.255.0]: >> + echo vr0 : 192.168.0.1/255.255.255.0 >> vr0 : 192.168.0.1/255.255.255.0 >> >> >> And this is what I see with 8: >> >> + DEFINT=vr0 >> + DEFIP=192.168.0.1 >> + DEFMASK=255.255.255.0 >> + read -p Enter network interface [vr0]: -t 5 INT >> Enter network interface [vr0]: >> + read -p Enter IP address [192.168.0.1]: -t 5 IP >> Enter IP address [192.168.0.1]: >> + read -p Enter netmask [255.255.255.0]: -t 5 MASK >> Enter netmask [255.255.255.0]: >> /: cho >> /: >> >> Strange that the "echo" statement is missing the first "e" character in the debug output. >> >> Even stranger on 8-CURRENT, if I *do* enter input before the newline (say I change the IP address or the network interface), the first character is not echoed back to the screen but is still saved to the variable. Here's an example when I run the script and provide input at each prompt: >> >> + DEFINT=vr0 >> + DEFIP=192.168.0.1 >> + DEFMASK=255.255.255.0 >> + read -p Enter network interface [vr0]: -t 5 INT >> Enter network interface [vr0]: r0 >> + read -p Enter IP address [192.168.0.1]: -t 5 IP >> Enter IP address [192.168.0.1]: 92.168.0.1 >> + read -p Enter netmask [255.255.255.0]: -t 5 MASK >> Enter netmask [255.255.255.0]: 55.255.255.0 >> + echo br0 : 192.168.0.1/255.255.255.0 >> br0 : 192.168.0.1/255.255.255.0 >> + echo ifconfig br0 inet 192.168.0.1 netmask 255.255.255.0 >> >> Notice that when I'm prompted, the first character doesn't echo but is still saved in the variable. >> >> >> Does anyone else see this same behavior? Any ideas on how to reset it back to how it works in STABLE? I'm not doing anything special with IFS so I'm stumped on how to troubleshoot this. >>>How-To-Repeat: >> use the above script in 8-CURRENT and validate output when Enter/Return is hit on the keyboard when waiting for default values >>>Fix: >> >> unknown >> >> >>>Release-Note: >>>Audit-Trail: >>>Unformatted: > > That is indeed a problem. Usually one EOL char would terminate read > but now it's two. > > I'm reproducing this issue on an older version of CURRENT (~2.5 weeks > old). I'll try looking through the commit logs for anything fishy... > > -Garrett None of the changes in .../src/bin/sh are causing this (I rolled back all files individually changed in the past 2 years and I couldn't find any problem files). It's most likely related to the MPSAFETTY changes. It'd be nice to use a livefs CD from before the MPSAFETTY changes were merged in to verify this. -Garrett