103

Unix/Linux 网络时间协议版本3 Network Time Protocol Version 3 (NTP) distribution for Unix systems

Received: from snow-white.ee.udel.edu by whimsy.udel.edu id aa22801;
          11 Mar 97 20:23 GMT
Received: by carbon.chem.nyu.edu (SMI-8.6/1.20)
	id PAA29481; Tue, 11 Mar 1997 15:19:08 -0500
Date: Tue, 11 Mar 1997 15:19:08 -0500
Message-Id: <[e-mail]>
From: "Edward J. Huff" <[e-mail]>
To: [e-mail]
CC: [e-mail], [e-mail], [e-mail], 
    [e-mail], [e-mail]
In-reply-to: <[e-mail]> ([e-mail])
Subject: Re: Adding volatile to variables shared with signal handlers

>Thanks for the alert. I suspect we have not heard the last of this
>issue.

I said "volatile required for variabled modified by signal handler."
Actually, it is possible to construct a case where volatile would be
required to force the compiler to actually store the values specified
into the static variables in the mainline routine so that the signal
handler will read the correct value.  So I suppose strictly speaking
the variables which are read by the signal handler should also be
marked volatile. 
 
activefds        INPUT ONLY
maxactivefd      INPUT ONLY
refio            INPUT ONLY
ninterfaces      INPUT ONLY
debug            INPUT ONLY

Actually, the book says "You specify volatile qualified types for
objects accessed or altered by signal handlers, by concurrently
executing programs, or by special hardware (such as memory-mapped
I/O control register)."

Also "A volatile qualified type indicates that agencies unknown to the
translator can access or alter the value stored in the object.  The
translator can assume that it has complete control of all objects that
do not have volatile qualified types."