how to put a shell call in a C/C++ program?

Hano · 04-27-2002, 04:14 PM

Hi everyone!

anybody ever tried to put a shell call from a C/C++ program?
,say, i would like to put a mkdir in it, or a pwd call and get the string back on the program...

Hano

acid_kewpie · 04-27-2002, 04:18 PM

you want the system() function.

gui10 · 04-28-2002, 11:20 PM

i think u can try reading up and trying some stuff from the unistd.h library

Hano · 05-09-2002, 02:14 PM

Ok, but now suppose you want to automate a connection to a host with ssh, do some stuff like read some file or exec something, i would guess that

system(" ssh login@host.org");
system(" ls -l"); // i would guess this do the ls on the remote host

but occurs that the ssh calls doesn't continue until you have ended the connection, and the ls -l command is executed after so the question is:

How can you automate a system call to be executed on the ssh client?

Hano

[cacheflow] · 05-09-2002, 04:38 PM

Quote:

Originally posted by acid_kewpie
you want the system() function.

maybe but its a very bad idea to use it... its poorly optimized and slow as hell...

Code:

#include <process.h>
#include <stdio.h>

int main(void)
{
   spawnl(P_WAIT, "myprog.exe", "myprog.exe", NULL);
   return 0;
}

Hano · 05-10-2002, 02:06 PM

Yes, if you allow me the objection that process.h and spawnl() doesn't belong to the linux enviroment, but to windoze

Hano

Hano · 05-10-2002, 02:18 PM

But more on a more concrete footing, what i'm looking for is a way to call a ssh, somehow return the port / terminal in with it has opened, and then write and read to the port / terminal from a C / C++ program, just like the opening a istream / ofstream ....

There HAVE to be an easy way to do this, Isn't linux supposed to be all about this?

Hano

Hano · 05-10-2002, 03:15 PM

I'm checking out a little about CORBA, may be i have to go all that way down.....

Hano

kervin · 05-19-2002, 02:26 PM

Be very careful with using ssh in that manner, your program will be prone to security holes. Eg. if a user's PATH environment variable gets changed, or LDLIBRARY_PATH, etc.

ssh allows you to run programs on the remote machine like rsh, eg. "ssh myserver.com ls -al /etc". anything after the server name would be treated as the remote command. If you have public key auth configured even better, then your program would not have to interact with the user, and you would not have to "write" to the ssh port ( I'm not sure thats possible at anyrate ).

Another, even better alternative is to use the OpenSSL library to encrypt your own connection to the remote server. That sounds like it's closer to what you want to do. OpenSSL has a set of well documented API calls to make a simple SSL or TLS connection. After this, you can treat the connection as a normal socket connection. http://openssl.org/