The problem is not obvious, both duff and Roger have invetigated, and they are experienced. It has to be something else. Why don't you debug more yourself under your very own environment, with your real data. Here I gave you something to start with: (code has only been syntax checked with -c)

my $newUser ='dummy';
while(<SETORA>)
{
    print "read from file: [$_]\n";
    if (/O_PREFIX=/) {
        print "regexp matched\n";
        s/O_PREFIX=\w+/O_PREFIX=$newUser/;
        print "after s, [$_]\n";
        print outputfile 
    }
}
[download]

I don't think your data really match the regexp's you have.

Any number of things may be wrong with your code. You haven't given us enough information to tell. You don't even say what output you are getting instead of what you want. Here are a few ideas of what could be wrong:

• Not really a problem but the conditional is redundant. The s/// will only do its job if the LHS actually matches.
• is outputfile really a filehandle opened for output?
• Is SETORA really a filehandle opened for intput?
• Are you sure you're checking the right file to see your results?

duff , sorry if my question was not clear , ok lets put it this way what if I want to just do : my input fils has only the following

    export O_PREFIX=OLD
[download]

I am doing the follwoing to change the value from OLD to NEW:

my $newValue='NEW';
s/O_PREFIX=(\w+)/O_PREFIX=$newValue/ if /O_PREFIX=/;
[download]

my code is working and I am reading from the correct file , however my output looks like this

O_PREFIX=NEWOLD
instead of 
O_PREFIX=NEW
[download]

I hope this make it clear , if not then sorry :(

use strict;
use warnings;

my $newUser ='$dummy';
while (<DATA>) {

    s/O_PREFIX=\w+/O_PREFIX=$newUser/;

    # s/(?<=O_PREFIX=)\w+/$newUser/;

    print;
}

__DATA__
if [ $YOU = 'frog' ]
   then
     export OUSER=TNN
     export CAR=SWs
     export O_PREFIX=ss
     export OS=corvette

fi
[download]

if [ $YOU = 'frog' ]
   then
     export OUSER=TNN
     export CAR=SWs
     export O_PREFIX=dummy
     export OS=corvette

fi
[download]

\w usually matches any of A-Z, a-z, 0-9, or _. If you have a larger character set you need to match, \w is not the right choice. Perhaps you could try it ass/O_PREFIX=.*/O_PREFIX=$newUser/ (which would replace from = up to the end of the line) or s/O_PREFIX=\S+/O_PREFIX=$newUser/ (which would replace from = up to a whitespace character (space, formfeed, tab, linefeed, or carriage return)).

(What \w matches may be expanded if in the scope of use locale or matching on Unicode data.)

Perhaps you can show us some of that code.