From: cgf AT bbc DOT com (Christopher Faylor)
Subject: Re: rxvt and latest Ian's changes.
2 Feb 1998 11:16:45 -0800
Message-ID: <EnrMs4.5oB.cygnus.cygwin32.developers@bbc.com>
References: <EnrL1L DOT 52q AT bbc DOT com>
Reply-To: gnu-win32-developers AT cygnus DOT com
To: cygwin32-developers AT cygnus DOT com

In article <199802021842 DOT NAA11902 AT subrogation DOT cygnus DOT com>,
Ian Lance Taylor  <gnu-win32-developers AT cygnus DOT com> wrote:
>   From: cgf AT bbc DOT com (Christopher Faylor)
>   Date: Mon, 2 Feb 1998 18:36:56 GMT
>
>   Could the process which has the slave open signal EOF in some other
>   way, perhaps, like by releasing a mutex?
>
>Not easily.  The problem is that the master will call read, which will
>turn into a call to ReadFile.  That will hang reading from a pipe.
>When all instances of the write end of the pipe are closed, the read
>will return correctly.
>
>Using a mutex is hard because Windows, in its wisdom, does not permit
>asynchronous reads from a pipe.  So as far as I know this would
>require creating a thread to hang reading from the pipe, and having
>the main code wait for either the thread or for the EOF mutex.  It
>might be possible to make this work, but what a pain just to do a
>read.  We would still have to keep track of the number of slaves ptys
>which were open, to know whether we should presume an EOF or not.

It does sound like too much bookkeeping for such a simple concept.

The other thing that I always investigate and then toss out is to use
mail slots for this type of application.  I don't think mail slots are
full-featured enough either and there may be some performance concerns.
-- 
http://www.bbc.com/	cgf AT bbc DOT com			"Strange how unreal
VMS=>UNIX Solutions	Boston Business Computing	 the real can be."