Re: [PATCH 2/2] daemon: graceful shutdown of client connection

From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: Apr 13, 2018, 13:03 UTC
Message-ID: <nycvar.QRO.7.76.6.1804131440100.65@ZVAVAG-6OXH6DA.rhebcr.pbec.zvpebfbsg.pbz>
In-Reply-To: <20180412210757.7792-3-kgybels@infogroep.be>

Hi Kim,

On Thu, 12 Apr 2018, Kim Gybels wrote:

Show 6 quoted lines

> On Windows, a connection is shutdown when the last open handle to it is
> closed. When that last open handle is stdout of our child process, an
> abortive shutdown is triggered when said process exits. Ensure a
> graceful shutdown of the client connection by keeping an open handle
> until we detect our child process has finished. This allows all the data
> to be sent to the client, instead of being discarded.

Nice explanation!

Show 6 quoted lines

> @@ -928,13 +931,13 @@ static void handle(int incoming, struct sockaddr *addr, socklen_t addrlen)
>  	}
>  
>  	cld.argv = cld_argv.argv;
> -	cld.in = incoming;
> +	cld.in = dup(incoming);

At first I was worried that somebody might want to remove this in the future, but then I saw this line (which also calls dup()):

Show 10 quoted lines

>  	cld.out = dup(incoming);
>  
>  	if (start_command(&cld))
>  		logerror("unable to fork");
>  	else
> -		add_child(&cld, addr, addrlen);
> +		add_child(&cld, addr, addrlen, incoming);
>  }
>  
>  static void child_handler(int signo)

Nice work!

I wonder whether you found a reliable way to trigger this? It would be nice to have a regression test for this.

Ciao, Dscho