Too many open files ... Why so? I looked at Kelp::Request sources, and it just transfers the filehandle for Plack. Why Plack does not closes it after rendering the requests?

Keep looking, don't stop, trace the code

Or just report it to the developers

Hi , see sub response_cb in Plack::Util ... it closes handles

I don't doubt basiliscos runs out of handles, I just think its his code that isn't closing them

Let me clarify the question: who is responsible for closing the filehandle? My code, framework (Kelp) or Plack?

Let me clarify the question: who is responsible for closing the filehandle? My code, framework (Kelp) or Plack?

Everybody is responsible :)

You assume the problem is with Kelp or Plack but you haven't ruled out your own code; you always have to rule out your own code first

If you can write a minimal Kelp app serving files that runs out of filehandles, only then can you assume there is a problem with Kelp or Plack, until then you should assume the problem is with your code

The problem is not with Kelp::Response::render_binary as it doesn't leave handles open.

It hands the handle to Plack::Response::body, and eventually sub response_cb in Plack::Util which reads from the handle and then closes the handle.

 

On a side note, if don't already know, you should ignore flushells the troll

Questions of this nature can be pretty-much categorically answered:   “it must be You!”

• “Plack is a plumber,” solely responsible for getting a request delivered from Apache to you, and a subsequent response delivered from you to Apache.   No smoking-gun here...
• Kelp, likewise, is a framework:   “generically” engineered to taking care of “whatever is ‘generic,’” so that you don’t have to.
• ... which leaves:   you.   Your application must be prepared to handle many (thousands of?) requests without “leaking.”   (It can, to a certain extent, “take the easy way out” by asking the PSGI manager to arrange for you to “commit hari-kiri” after so-many requests, but that is a crutch.)   You should always assume that you are responsible for closing all resources.

my $ih = $image->filehandle('<', ':raw');
my $ret = $self->res->set_content_type('image/x-icon')
          ->render_binary($ih);
close $ih;
return $ret;
[download]

Update: Rethinking this, the handle created by $image->filehandle and passed to render_binary() is being stored in some object. So, when that object gets destroyed, the handle should be automatically closed.

Yes, that would close the filehandle before the file was sent, so no file would be sent

See Re^2: plack: close filehandle responsibility (everybody), while https://metacpan.org/pod/Plack::Response#body doesn't document that it closes $io the handle it accepts as argument, it does do it

Try it out, fireup this plack program, and start taskmanager, and watch the handles count not increase with each browser request you make

#!/usr/bin/perl --
use strict;
use warnings;

use Plack::Builder;
use Path::Tiny qw/ path /;

my $app = sub {
    return [
        200,
        [
            'Content-Type' => 'text/plain; charset=UTF-8',
        ],
        path( __FILE__ )->openr_raw,
    ];
};

my $finalapp = builder {
  enable 'StackTrace';
  $app;
};

if( $0 eq __FILE__ ){ 
    require Plack::Runner;
    my $runner = Plack::Runner->new;
    $runner->parse_options( qw' --host 127.0.0.1 --port 5000 ' );
    $runner->run( $finalapp ); ## perl      this.psgi
} else {
    $finalapp; ## plackup -l localhost:5000     this.psgi
}
## lwp-request -Ed http://127.0.0.1:5000/
[download]

If I close the handle openr_raw returns before Plack reads it without slurping, there will probably be a plack error

update: I just tested it, no error, but no file is sent either as expected

I don't see where your code (in your response to my post) is calling render_binary().