Re: mod_perl and CGI behavior

[davebaker]

Haven't tested this yet as to whether it makes a difference, but I see you've imported various "standard" functions from CGI.pm by including :standard in the "use CGI qw(:standard);" statement. Because you're using CGI.pm in its object-oriented style (my $cgi = CGI->new() etc.), you don't need to do that, and you might be better off with "use CGI;" instead.

[naChoZ]

Unfortunately it seems deeper than that.

Looking at the code over on github here, you can see that the ->new method checks if mod_perl is being used and, if it is, gives you back an object with the request attached to it, and all that stuff...

But, regardless, I dumbed down the script to just...

#!/perl/bin/perl

use strict;
use warnings;

use Data::Dumper::Names;
use CGI;

use Apache2::Connection ();
use Apache2::RequestRec ();


$| = 1;

my $header = CGI::header();

print STDERR Dumper($header);

...and it still outputs just...

$header = '';

[davebaker]

Yes, something going on at line 1617 of CGI.pm at github, inside sub "header" --

if (($MOD_PERL >= 1) && !$nph) {
        $self->r->send_cgi_header($header);
        return '';
}

which explains the empty set being returned, after it's invoked a "send_cgi_header" mod_perl method that seems to be designed to take the place of sub "header" as a practical matter.

[naChoZ]

Interesting. That did the trick actually. So this test code...

     my $test_header = $cgi->header(
         -type => 'application/xml'
     );

     ddump('test_header', __LINE__, $test_header);

     my $another_test_header = $cgi->header(
         -nph => 1,
         -type => 'application/xml'
     );

     ddump('another_test_header', __LINE__, $another_test_header);

...produced the following debug output...

$test_header_line_673_1 = '';
$another_test_header_line_680_1 = 'HTTP/1.1 200 OK
Server: Apache/2.4.34 (Red Hat) OpenSSL/1.0.2k-fips mod_perl/2.0.11 Pe
+rl/v5.22.4
Date: Wed, 07 Oct 2020 21:53:46 GMT
Content-Type: application/xml; charset=ISO-8859-1

';

[naChoZ]

Interesting. I wonder what that nph (no parsed header) business ends up doing... I'll have to experiment with that.

Thanks!