www-apache-bugdb mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Brian Behlendorf <br...@hyperreal.org>
Subject Re: mod_cgi/2127: Impossible to detect browser disconnect from CGI
Date Sun, 07 Jun 1998 17:00:01 GMT
The following reply was made to PR mod_cgi/2127; it has been noted by GNATS.

From: Brian Behlendorf <brian@hyperreal.org>
To: rz10@cornell.edu (Ray Zimmerman)
Cc: apbugs@apache.org
Subject: Re: mod_cgi/2127: Impossible to detect browser disconnect from
Date: Sun, 07 Jun 1998 09:57:24 -0700

 At 11:18 AM 6/5/98 -0400, Ray Zimmerman wrote:
 >I just got back from a month of vacation ... I was wondering about the
 >status of this report. I see you changed it from "open" to "analyzed" and
 >was wondering what the conclusion of the analysis was.
 The analysis was that thread of discussion that went on about it, which was
 Cc'd to you I believe.  You can see it all at
 >Is it a bug?
 Well, it doesn't appear that folks seem to think it is, at least not a
 serious one.  As you can see, neither Marc (who's probably testing it on
 FreeBSD) or Lars (who tested Linux) can see this; they claim the CGI is
 killed fairly quickly if it writes an output and the client disappears.
 >If so, who's working on fixing it?  And can I be of help?
 The 75 seconds certainly is not anything in the Apache code; it could be
 that Apache doesn't detect the disconnect from the client for 75 seconds
 after you hit "stop" due to something in Solaris's TCP stack.  
 Look in the error logs.  Do you see entries like:
 [Sun Jun  7 09:46:22 1998] [info] client stopped connection
 before send body completed
 Well you should (given you're using a low-enough loglevel setting),
 everyone does, that's the sign that Apache has been told the connection has
 been closed, and the CGI script should have been killed right after that.
 If Apache doesn't get that signal for 75 seconds after the "stop" button
 had been hit, there's not much that it can do.  If you're saying there's 75
 seconds (or however many) between the time you see that entry in the error
 logs, there's an error in Apache.  But otherwise, it sounds like a Solaris
 and TCP issue...
 pure chewing satisfaction                                  brian@apache.org

View raw message