Escolar Documentos
Profissional Documentos
Cultura Documentos
Page 1 of 4
The Problem
The problem I always encountered in scripting ftp transfers involved getting a password to the ftp server. Typical ftp client programs under Unix, Linux, Solaris and NetBSD all read the ftp password from /dev/tty.
The above script will just hang if run in the foreground (in an xterm), or if run in the background (from a cron job), it will fail to perform the work of transferring file.txt.
/dev/tty names a strange, magic device. Each process (more strictly each process group) has a different /dev/tty, and you can not naively make ftp clients read the password from some non-
magic, yet convenient source, like a "here document". When run in an xterm, the script above appears to hang because it reads the password from /dev/tty. The xterm constitutes the script's /dev/tty, so the script waits for keyboard input.
http://www.stratigery.com/scripting.ftp.html
25-04-2013
Page 2 of 4
FILE='file.txt' ftp -n $HOST <<END_SCRIPT quote USER $USER quote PASS $PASSWD put $FILE quit END_SCRIPT exit 0
The Tricks
Getting the password to the ftp server without having the ftp client program read the password from /dev/tty requires two tricks: 1. Using the -n option on the ftp client program to prevent the ftp client from trying to log in immediately. That way, the ftp client does not ask for a user ID and password. No use of/dev/tty. 2. Use the ftp client program command quote to send user ID and password to the ftp server.
Further Refinements
The above sh script will spew lots of ftp client output to standard output. Even if everything works perfectly, the user running the above script will see lots of incomprehensible text scrolling by quite rapidly. One refinement would send output to different places:
ftp -n $HOST > /tmp/ftp.worked 2> /tmp/ftp.failed <<END_SCRIPT
One could further refine error handling by acting on the ftp client program's exit status:
ftp -n $HOST > /tmp/ftp.worked 2> /tmp/ftp.failed <<END_SCRIPT blah blah END_SCRIPT EXITSTATUS=$? if [ $EXITSTATUS != "0" ] then # handle the error... fi
Except that the above doesn't always work - most FTP clients always exit with a status of 0. This leads to ugly "false negatives": the file transfer fails, but the script doesn't detect the problem. One way to verify that a file transfer took place - transfer it back:
#!/bin/sh ftp -n << END_SCRIPT open $1 user $2 $3 put $4 get $4 retrieval.$$ bye END_SCRIPT
http://www.stratigery.com/scripting.ftp.html
25-04-2013
Page 3 of 4
if [ -f retrieval.$$ ] then echo "FTP of $4 to $1 worked" rm -f retrieval.$$ else echo "FTP of $4 did not work" fi
Regular FTPs there and back of large files can consume a lot of time.
Alternative #1
I saw a second way of doing this in a usenet article:
#!/bin/sh USER=userid PASSWD=userpw ftp -n f2dev <<SCRIPT user $USER $PASSWD binary get some.file quit SCRIPT
It still uses the "-n" trick, but it sends user ID and password in the same "user" command.
Alternative #2
Use a .netrc file Linux, Unix and BSD users have the alternative of using a .netrc file. The ftp man page documents the format of .netrc. To accomplish the task of using ftp in a shell script you would have to fill out a .netrc file something like this:
machine something.else.com login myid password mypassword
ftp demands that .netrc not have group or world read or write permissions:
$ ls -l .netrc -rw------1 bediger
users
Using a .netrc file has a few problems that may or may not prevent you from using it.
http://www.stratigery.com/scripting.ftp.html
25-04-2013
Page 4 of 4
A shell scripkt that does FTP using .netrc is no longer self-contained. You have to keep track of two files, which means that bugs can be less than obvious. ftp reads it's user ID's .netrc. If you develop your script under a given user ID, then put it in production under a second user ID, you have to coordinate .netrc file contents between those two user IDs.
Alternative #3
Apparently, the Ckermit program from Columbia University understands FTP. You could use Ckermit to script FTP transfers. This looks to have advantages and disadvantages. On the "pro" side, it appears that Ckermit can exit on various errors, like unknown user IDs, or bad passwords. On the "con" side, you have to have Ckermit. I don't recall that it had a too onerous install, but it doesn't come with many Linux distros these days, and it probably doesn't come with any vendor Unix. $Id$
http://www.stratigery.com/scripting.ftp.html
25-04-2013