Unfortunately CGI.pm has been removed from core Perl, but there are still lots of applications out there that use CGI.pm. I even encountered one built in 2017 by people who have not learned newer techniques. The thing is, CGI is perfectly good for small things and for trying out new concepts without forcing the developer to learn some new technology.

Anyway, when I got to this client, I had to help them write tests for their CGI-based application.

I did not want to convert the application to some PSGI-based application, even though that could be a nice route. We just wanted to write some tests for the application.

Hence we used the command-line capabilities of CGI.pm.

This is a sample CGI script:


use strict;
use warnings;

use CGI;

my $q = CGI->new;
print $q->header;

print qq{<html><head></head><body>\n};
print qq{<table>\n};
printf qq{<tr><td>request_method</td><td>%s</td></tr>\n}, $q->request_method;
for my $p ($q->param()) {
    printf qq{<tr><td>%s</td><td>%s</td></tr>\n}, $p, scalar $q->param($p);
print qq{</table></html>\n};

It does not do much, but it is enough to show how to test any CGI-script.


use strict;
use warnings;

use Test::More;
use Capture::Tiny qw(capture);
use File::Temp qw(tempdir);

plan tests => 2;

subtest get => sub {
    plan tests => 5;

    local $ENV{REQUEST_METHOD} = 'GET';
    local $ENV{QUERY_STRING}   = 'name=foo&email=bar@corp.com';

    my ($out, $err, $exit) = capture { system "./cgi.pl" };

    #diag $out;
    like $out, qr{<tr><td>request_method</td><td>GET</td></tr>}, 'GET';
    like $out, qr{<tr><td>name</td><td>foo</td></tr>}, 'name';
    like $out, qr{<tr><td>email</td><td>bar\@corp.com</td></tr>}, 'email';
    is $err, '', 'stderr is empty';
    is $exit, 0, 'exit code is 0';

subtest post => sub {
    plan tests => 5;

    my $params = 'language=Perl&creator=TimToady';
    local $ENV{REQUEST_METHOD} = 'POST';
    local $ENV{CONTENT_LENGTH} = length($params);

    my $dir = tempdir(CLEANUP => 1);
    my $infile = "$dir/in.txt";
    open my $fh, '>', $infile or die "Could not open '$infile' $!";
    print $fh $params;
    close $fh;

    my ($out, $err, $exit) = capture { system "./cgi.pl < $infile" };

    #diag $out;
    like $out, qr{<tr><td>request_method</td><td>POST</td></tr>}, 'POST';
    like $out, qr{<tr><td>language</td><td>Perl</td></tr>}, 'language';
    like $out, qr{<tr><td>creator</td><td>TimToady</td></tr>}, 'creator';
    is $err, '', 'stderr is empty';
    is $exit, 0, 'exit code is 0';

There are two test-cases. One for a GET request and one for a POST.


For the GET request we need to put the parameters that you'd see in the URL into the QUERY_STRING environment variable. In addition we need to set the REQUEST_METHOD to GET. Then we can run our CGI script. We use ./cgi.pl instead of $^X cgi.pl or even just perl cgi.pl as this will ensure we use the same perl for the execution of the CGI script as it is used by the server.

We use the capture function provided by Capture::Tiny to capture the STDOUT and STDERR of the program. The STDOUT should contain everything that is sent back to the browser, including both the header and the body, but not the HTTP status.

The STDERR should be empty. Whatever goes there usually ends up in the error-log of the web server, but normally that should be empty.

The 3rd value returned by the capture function is the exit-code of the system call in the block. Normally it should be 0, indicating success.

Once we have the responses, we can use the regular functions of Test::More such as is and like to compare the result to the expected values or to check if the result contains something we expect.


In the case of a POST request we need to set the REQUEST_METHOD to be POST, we need to supply the input on the STDIN of the CGI-script and we need to supply the length of the input in the CONTENT_LENGTH environment variable. For this to work nicely, we create a temporary directory using the tempdir function of File::Temp and in the system call we set up redirection of the STDIN of the cgi.pl execution.


local is used when setting the environment variables as that, and the curly-braces around the subtest functions will prevent these settings from leaking between the subtests.

Running the test

Using prove we can run the test script and get a very compact output:

$ prove cgi.t

cgi.t .. ok
All tests successful.
Files=1, Tests=2,  1 wallclock secs ( 0.04 usr  0.01 sys +  0.13 cusr  0.04 csys =  0.22 CPU)
Result: PASS

Further reading

There are two series of related articles. One is about testing with Perl and the other one is about CGI and Perl.

Besides the 3 environment variables we have already seen there are a few others that might need to be set in order to fully recreate the environment provided by the web server. The following ones are the most common variables:

local $ENV{QUERY_STRING}    = 

local $ENV{HTTP_COOKIE}     =
local $ENV{HTTP_HOST}       =
local $ENV{HTTP_REFERER}    =
local $ENV{PATH_INFO}       =