Problems looping to prompt for another password - tcl

Question

I need some help with an EXPECT script please....
I'm trying to automate a login, prior to accessing a load of hosts, and cater for when a user enters a password incorrectly. I am getting the username and password first, and then validating this against a particular host. If the password is invalid, I want to loop round and ask for the username and password again.
I am trying this :-
(preceding few irrelevant lines omitted)
while {1} {
send_user "login as:- "
expect -re "(.*)\n"
send_user "\n"
set user $expect_out(1,string)
stty -echo
send_user "password: "
expect -re "(.*)\n"
set password $expect_out(1,string)
stty echo
set host "some-box.here.there.co.uk"
set hostname "some-box"
set host_unknown 0
spawn ssh $user#$host
while {1} {
expect {
"Password:" {send $password\n
break}
"(yes/no)?" {send "yes\n"}
"Name or service not known" {set host_unknown 1
break}
}
}
if {$host_unknown < 1} {
expect {
"$hostname#" {send "exit\r"
break
}
"Password:" {send \003
expect eof
close $spawn_id
puts "Invalid Username or Password - try again..."
}
}
} elseif {$host_unknown > 0} {
exit 0}
}
puts "dropped out of loop"
And now I can go off and do lots of stuff to lots of boxes .....
This works fine when I enter a valid username or password, and my script goes off and does all the other stuff I want, but when I enter an invalid password I get this :-
Fred#Aserver:~$ ./Ex_Test.sh ALL
login as:- MyID
password: spawn ssh MyID#some-box.here.there.co.uk
Password:
Password:
Invalid Username or Password - try again...
login as:- cannot find channel named "exp6"
while executing "expect -re "(.*)\n""
invoked from within "if {[lindex $argv 1] != ""} {
puts "Too many arguments"
puts "Usage is:- Ex_Test.sh host|ALL"
} elseif {[lindex $argv 0] != ""} {
while {1} {
..."
(file "./Ex_Test.sh" line 3)
Its the line "can not find channel named "exp6" which is really bugging me.
What am I doing wrong? I am reading Exploring Expect (Don Lines) but getting nowhere....

Answer 1

Whenever expect is supposed to wait for some word, it will save the spawn_id for that expect process into expect_out(spawn_id).
As per your code, expect's spawn_id is generated when it encounters
expect -re "(.*)\n"
When user typed something and pressed enter key, it will save the expect's spawn_id. If you have used expect with debugging, you might have seen the following in the debugging output
expect does "" (spawn_id exp0) match regular expression "(.*)\n"
Lets say user typed 'Simon', then the debugging output will be
expect: does "Simon\n" (spawn_id exp0) match regular expression "(.*)\n"? Gate "*\n"? gate=yes re=yes
expect: set expect_out(0,string) "Simon\n"
expect: set expect_out(1,string) "Simon"
expect: set expect_out(spawn_id) "exp0"
expect: set expect_out(buffer) "Simon\n"
As you can see, the expect_out(spawn_id) holds the spawn_id from which it has to expect for values. In this case, the term exp0 pointing the standard input.
If spawn command is used, then as you know, the tcl variable spawn_id holds the reference to the process handle which is known as the spawn handle. We can play around with spawn_id by explicitly setting the process handle and save it for future reference. This is one good part.
As per your code, you are closing the ssh connection when wrong password given with the following code
close $spawn_id
By taking advantage of spawn_id, you are doing this and what you are missing is that setting the expect's process handle back to it's original reference handle. i.e.
While {1} {
###Initial state. Nothing present in spawn_id variable ######
expect "something here"; #### Now exp0 will be created
###some code here ####
##Spawning a process now###
spawn ssh xyz ##At this moment, spawn_id updated
###doing some operations###
###closing ssh with some conditions###
close $spawn_id
##Loop is about to end and still spawn_id has the reference to ssh process
###If anything present in that, expect will assume that might be current process
###so, it will try to expect from that process
}
When the loop executes for the 2nd time, expect will try to expect commands from the spawn_id handle which is nothing but ssh process which is why you are getting the error
can not find channel named "exp6"
Note that the "exp6" is nothing but the spawn handle for the ssh process.
Update :
If some process handle is available in the
spawn_id, then expect will always expect commands from that
process only.
Perhaps you can try something like the following to avoid these.
#Some reference variable
set expect_init_spawn_id 0
while {1} {
if { $expect_spawn_id !=0 } {
#when the loop enters from 2nd iteration,
#spawn_id is explicitly set to initial 'exp0' handle
set spawn_id $expect_init_spawn_id
}
expect -re "(.*)\n"
#Saving the init spawn id of expect process
#And it will have the value as 'exp0'
set expect_init_spawn_id $expect_out(spawn_id)
spawn ssh xyz
##Manipulations here
#closing ssh now
close $spawn_id
}
This is my opinion and it may not be the efficient approach. You can also think of your own logic to handle these problems.

Answer 2

You simply need to store the $spawn_id as a temp variable before a nested expect command, then set the $spawn_id to the temp variable after a nested expect command.
Also, get rid of the while {1} loops. They are not needed because expect behaves like a loop provided you use exp_continue whenever you don't wish to exit. You don't need expect eof nor do you need close $spawn_id. I don't use them in the following example:
#!/usr/bin/expect
set domain [lindex $argv 0];
set timeout 300
spawn ./certbot-add.sh $domain
expect {
"*replace the certificate*" {
send "2\r";
exp_continue;
}
"*_acme-challenge*" {
puts [open output.txt w] $expect_out(buffer)
spawn ./acme-add.sh $domain
set tmp_spawn_id $spawn_id
expect {
"$ "
}
set spawn_id $tmp_spawn_id
send "\r";
exp_continue;
}
"*certificate expires on*" {
puts "Certificate Added!"
}
}