Suggested Steps to Implement ISH

First Part

• Decide the interface between your parser and rest of your program. You may make it simple at first and modify it as you proceed, or you may want to have the data structures defined at the very beginning. In your final version, the commands you have to process looks like this:

ish> command1 arg1 arg2 <infile >outfile

ish> command1 arg1 arg2 <infile | command2 arg1 arg2 arg3 | command3 arg1 arg2 arg3 >outfile

Refer to Command Line Parsing.

• Write your parser to parse simple commands without handling redirection and & (i.e.: ish> command arg1 arg2 arg3)
• Use fork() and execvp() to execute the command given, report error when execution does not succeed. (execvp() will search the PATH environment variable for the command for you. You are NOT allowed to use this in the code you turn in; you must use execve() by then.) Make sure your shell will also work if its input is redirected. (i.e.: ish <.ishrc)

Sample test.

• Make your shell read and execute the .ishrc file in your current directory when it starts; you don't have to report an error if that file does not exist. (From now on, you are assumed to report errors whenever appropriate.) Your shell should enter normal interactive mode after executing .ishrc unless .ishrc contains an exit command (then you shell can exit immediately after seeing this command).
• Implement built-in commands cd and exit

Sample test.

• Implement environment variables. You can use your code in Assignment 4.
• Implement bulit-in commands printenv, setenv, and unsetenv.

Sample test.

• Change your execvp() to execve(), search PATH environment variable for the command, and pass correct environment variables to execve(). Make sure your shell will work if the user specifies the absolute path of the command (e.g.: /usr/bin/date rather than date).

Sample test.

• Make your parser recognize redirection.
• Implement I/O redirection.

Sample test.

• Make your parser recognize pipelines.
• Implement pipelines without worrying about redirection combined with pipeline. Remember you have to job control in the near future, so your data structure should be convenient for you to handle all the processes in one pipeline easily.
• Make your shell work when pipeline and I/O redirection are combined (Only the first command in the pipeline can have its input redirected, only the last command in the pipeline can have its output redirected. Report an error if redirection in the pipeline does not fall into the above categories.)

Sample test.

Final test.

Second part

• Make your parser recognize &
• Implement background jobs which are started in the background using &. Use sleep 200 & and /usr/bin/ps -l -u user1 (substitute user1 with your userid) to check you did launch(its status in S column should be "S") and is running this job. At the same time type something at the prompt to make sure you can still do other things in the foreground.
• Make all processes in one command to be in one process group. Use the command(cat | wc | wc) and use /usr/bin/ps -j -u user1 to check the PGID of each process.
• Set up signal handlers. Especially, Ctrl+Z and Ctrl+C should be intercepted and handled by foreground job(not your ish). Use /usr/bin/ps -l -u user1 to check it. (The stopped processes will have a "T" in S(Status) column, all other processes should have "R" or "S" or "O" in there.) You might want to set up other signal handlers as well.
• Implement status reporting and assign each job appropriate job id. Report all jobs as either "Running" or "Stopped" now.
• Implement bg and fg
• Implement kill
• Modify your status reporting as the assignment homepage required
• This might not be a trivial task: Make sure your part1 still works after you add job control. If this task is quite difficult for you, you might want to make the structure of your programs better in the future.

Good luck.