The standard documentation appears for this under-appreciated command in the Expect extension here [L1 ].
exp_continue isn't strictly necessary. One could write a loop from other, more primitive, elements. However, exp_continue is a great convenience for the common situation where it can be used to easily implement various client/server models in expect, such as FTP, SMTP or HTTP (but see [L2 ]), where we are uncertain about the number and order of commands sent by server. [explain much more]
exp_continue is labeled as "under-appreciated" above because there's such a gap between the large number of people who can program basic exp_send-expect sequences, and the small minority who are familiar with exp_continue, despite the latter's usefulness.
Don Libes is quoted at http://coding.derkeiler.com/Archive/Tcl/comp.lang.tcl/2004-07/0437.html as saying "As far as the difference between while and exp_continue, it's mostly a matter of style. The exp_continue is also faster but it's imperceptible in most real-world applications."
[provide sample code for some client/server]
An example that uses exp_continue ( check if a user/password combination is valid ):
proc auth_user { user pass } {
global spawn_id
set success 0
log_user 0
spawn ssh -l $user localhost
set spid $spawn_id
expect -i $spid \
"word: " {
exp_send -i $spid "${pass}\r"
if { $success == 0 } {
incr success -1
exp_continue
}
} "> " {
exp_send -i $spid "exit\r"
set success 1
} "continue connecting (yes/no)? " {
exp_send -i $spid "yes\r"
exp_continue
} "incorrect" {
set success -4
exp_send -i $spid "\r"
} "try again." {
set success -4
exp_send -i $spid "\r"
} timeout {
set success -1
}
exp_close -i $spid
exp_wait
puts stderr authuser:$success
return [ expr ( $success == 1 ) ? 1 : 0 ]
}