From: ian AT cygnus DOT com (Ian Lance Taylor)
Subject: Re: Question
9 Apr 1998 16:02:13 -0700
Message-ID: <199804092244.SAA10273.cygnus.cygwin32.developers@subrogation.cygnus.com>
References: <01BD63A1 DOT EC97AE90 AT SOMESHIT>
To: sos AT prospect DOT com DOT ru
Cc: cygwin32-developers AT cygnus DOT com

   From: Sergey Okhapkin <sos AT prospect DOT com DOT ru>
   Date: Thu, 9 Apr 1998 10:26:13 +0400

   Cpp (as a first task in a pipeline) will be started with stdout redirected 
   to a pipe and with inheritted pipe's read end (in 
   last_pipe_input=pdes[READ_PORT])! If cc1 will terminate _before_ cpp 
   termination (due to fatal syntax error), cpp will hang in a write to 
   stdout.

You're right, that does seem to be a potential problem.

However, since the MS documentation about spawn says that files which
are open before the spawn call is made remain open in the spawned
process, I think that the cygwin32 of spawn should behave in the same
manner.  So I would say that this is a bug which should be fixed
somewhere in gcc, rather than in cygwin32.

Ian