From: Prakash Kailasa <PKailasa at seisint dot com>
To: Andrew Ford <A.Ford@ford-mason.co.uk>
Subject: cronolog patch
Date: Tue, 15 May 2001 19:35:41 -0400

Andrew,

I have been using cronolog for a few days. Thanks for developing
this very useful program and releasing it for public.

I have noticed a problem very frequently. On my development system,
we stop and start apache several times a day. Whenever I restart
apache using the 'graceful' option, one of the cronolog processes
(one associated with the error_log) is not terminating. After,
several graceful restarts, the cronolog processes are getting
accumulated, holding system resources. Eventualy I run out of
file descriptors.

The only recourse at this point is to stop apache completely and
start it again.

The enclosed patch fixes this problem. Please take a look at it
and see if it is okay to add it to your code.

Its possible the code is not very good. Or, there may be a better
way to do it. It's been nearly six years since I did any coding
in C.

Thanks,
/prakash

--- cronolog-1.6.1/src/cronolog.c.orig	Mon Dec 20 03:56:20 1999
+++ cronolog-1.6.1/src/cronolog.c	Tue May 15 19:09:43 2001
@@ -84,12 +84,14 @@
 
 #include "cronoutils.h"
 #include "getopt.h"
+#include <signal.h>
 
 
 /* Forward function declaration */
 
 int	new_log_file(const char *, const char *, mode_t, 
 		     PERIODICITY, char *, size_t, time_t, time_t *);
+void	terminate_self(int);
 
 
 /* Definition of version and usage messages */
@@ -243,6 +245,9 @@
     DEBUG(("periodicity = %s\n", periods[periodicity]));
 
 
+    /* set up signal handler to catch USR1 (graceful restart by apache) */
+    signal(SIGUSR1, terminate_self);
+
     /* Loop, waiting for data on standard input */
 
     for (;;)
@@ -350,4 +355,12 @@
 	create_link(pfilename, linkname, linktype);
     }
     return log_fd;
+}
+
+void terminate_self(int sig)
+{
+    time_t time_now = time(NULL);
+    DEBUG(("%s (%d): received signal USR1; terminating.\n",
+		timestamp(time_now), time_now));
+    exit(6);
 }