Shell Programming and Scripting

Hi,

I am writing a series of scripts for work to analyse intranet access logs. All of the scripts do as they should when run individually from the shell, but only when run from certain directories. This sounds like it may be a PATH issue but I am not sure. When I run a certain script, say "script.pl" from it's working directory - everything is fine. However, if I try to run the script from another location by specifying the full path name of the script, the script tries to execute and drops straight back to the shell without any output to the user. I do receive output if run from the working directory of the script.

The script is a Perl script, and I have the appropriate shebang line at the start of the file "#!/bin/perl". Also, the file has been CHMODed to execute.

Any help would be most appreciated.
Max

Thanks to anyone viewing my post. I seem to have solved the problem. I didn't think to export PATH in the ksh script that was calling my Perl scripts. I just added a PATH variable to the top of the script including any appropriate directories and exported that before I did anything else.

Thanks again.

EDIT: It is nice that you figured out what went wrong. My original response follows, which describes another possibility within perl (although it turned out that you don't fall in that case).

--------------------------------------------------------------------------------------

This is in fact a BIG question.

This is a symptom that the script is not well written because it relies on being invoked from a certain working directory. This is a common problem that many scripts may work on the command line but refuse to run with cron, for instance, as the working directory is different from norm.

There is no easy fix to this problem if the script is written without the sense of such relocation in mind. A quick fix is by some simple compound command which "cd" and then execute the script (that you can always write a script to automate these two steps).

If you want to fix this permanently, you must examine the script line by line for all references with relative paths. They may occur as two major classes:

1. Operations depending on the Perl include path
2. File open() and external shell commands

The Perl include path (@INC) holds the list of directories to search for libraries (used by "require" and "use"). Make sure the dirs listed cover all the paths needed by your require() statements. If any of the items listed denote relative path, change them to use absolute paths. Normally, @INC includes '.', so the current working directory is taken into account.

Other file operations also depend on current working directory in case of relative paths are used. Without '/' prefix, all paths are relative to CWD. For instance,

open FILE, ">>file.txt";
`diff file1.txt file2.txt`

These two classes of relative paths should generally be treated separately, because they are of different nature. Perl @INC is for resolving external Perl scripts, they should not normally be subject to changes in CWD. However, at other times you may actually wish to honour the CWD in certain cases. For instance, in a build script one may wish to have compiled object binaries placed outside of the original source tree, so we will change the CWD to outside the tree and have everything generated dropped in that directory.

However, in really big scripts nobody wants to hardcode an absolute path everywhere (for portability/management sake as well). It is therefore common practice to have a single file that acts as a "configuration file" that defines all these paths in a centralized manner. The @INC is initialized at the start of the script based on the contents of this file. But then, how to locate this file? Either it is placed at a fixed location, or you may actually deduce dynamically the base directory of the current script file with something like

use File::Spec;
dirname(File::Spec->rel2abs(__FILE__))

as an aid to construct absolute paths from relative paths. If all of the paths are constructed without implicit use of relative paths, your script will then be independent of the CWD.

Tedious enough. Isn't it?

That's fantastic - thank you for the very detailed response - always appreciated. If I have any further problems I will be sure to post...

Incidentally...

The next step I am having issues with is compressed files. For space reasons, the web logs that I am analysing are compressed nightly (they are typically in excess of 500Mb). If I wanted to uncompress these before performing any operations on them, would it just be a case of calling:

system(uncompress filename);

and then recompressing them in a similar fashion at the end? Or is it more involved fo you know?

Thanks for any help you can provide.