UNIX for Advanced & Expert Users

You could do it by parsing: There are only so many methods of invoking external commands:

$(command)
command | command
`command`

These constructs can be extracted from every script with some sed-script. Get the "command" part out of the found pieces and apply a "file" command on each. If the result is "executable" or "object module" or similar you can throw it away, if the result is "shell script" you have found one candidate.

This way you get a list of all the shell scripts a script invokes. Create a list for every script and you have your tree.

I hope this helps.

bakunin

Anyway,this is a good question which will be helpful in the analysis phase of a lot of enhancement projects.So i request you to post the solution you applied(either extracted from the replies or from some other ref - the working code).

With Regards
Dileep Pattayath

I don't think the idea I presented requires any manual work. Write short script to add the line I suggested to each script. Nothing manual there.

Parse the log file using command-line tools. Nothing manual there.

Of course, I could always be missing some important point you were trying to make, or I might not have been clear enough in my post. If the idea still doesn't work for you, please post the reason, or let us know what you end up doing.

Thanks,
ShawnMilo

Shawn,
You are assuming that all the scripts will be called in a reasonably short time frame (what about scripts only run at year end?)
There are also other issues that I have considered since my last post:
What about dynamic calls:
if .....
a=1
else
a=2
fi
abc$a

What about scripts called from within application code, and do you have the source code for this.

Jgt's point is valid one. What about scripts which are called from CRON or any other schedulers?

Hello All,

Fortunately in the system that I'm working on, I know all the scripts that are scheduled using the various scheduler apps, or are manually executed, so I know all the "starting points". None of the scripts of my concern get executed from the application code or called dynamically.

Shawn - I agree what you worte, this can be done automatially, but I just didn't want to modify a huge number of scripts for getting this information.

Under the circumstances, the solution I have opted for is a hybrid one (its not the only solution, and it may not be the best one) , composed of :
1. "trace", as I wrote in one of my earlier post - command:
trace -h -a -d -k $IGNORE -o /tmp/my_trace_log; trcon; SCRIPT_NAME ; trcoff; trcstop

and make the report by

trcrpt -O "exec=on,pid=on,timestamp=3" /tmp/my_trace_log > my_trace_log.rpt

$IGNORE practically contain ALL the categories mentioned in trchkid.h (see man page of trace) except those of FORK and EXEC. A lot of EXECs need to be grepped out from the final report, and I have what I'm looking for!

2. Another combination of solutions suggested above - I have certain scripts that I need to move from one directory to another and thus see what other scripts will be impacted due to hardcoded paths, so I wrote a fancy grepp to look through the whole filesystem limiting the searched files to "ascii text" files and "*Scripts*" (using command "file"). Sucked the life out of the box!

I have been able to pull out the information I required using the above approach.

I would like to emphasise that this solution may or may not work for all, and is certainly not covering all the scenarios. It does work for me, so I'm sharing it for everyone who took time to think about this problem.

Once again, thanks all!