Moved all of the background tasks back to the old EVT_TIMER style

[citadel.git] / citadel / sysdep.c

diff --git a/citadel/sysdep.c b/citadel/sysdep.c
index 9809cddbb4dbb56b6579c3a506b5e4eb368f5f41..4c19d2933586aed277a011fd3170fbc6cc18a442 100644 (file)
--- a/citadel/sysdep.c
+++ b/citadel/sysdep.c
@@ -168,22 +168,14 @@ void init_sysdep(void) {
         */
        sigemptyset(&set);
        sigaddset(&set, SIGINT);                // intr = shutdown
-       // sigaddset(&set, SIGQUIT);            // quit = force quit
        sigaddset(&set, SIGHUP);
        sigaddset(&set, SIGTERM);
-       // sigaddset(&set, SIGSEGV);            // we want core dumps
-       // sigaddset(&set, SIGILL);             // we want core dumps
-       // sigaddset(&set, SIGBUS);
        sigprocmask(SIG_UNBLOCK, &set, NULL);
 
        signal(SIGINT, signal_cleanup);         // intr = shutdown
-       // signal(SIGQUIT, signal_cleanup);     // quit = force quit
        signal(SIGHUP, signal_cleanup);
        signal(SIGTERM, signal_cleanup);
        signal(SIGUSR2, signal_exit);
-       // signal(SIGSEGV, signal_cleanup);     // we want coredumps
-       // signal(SIGILL, signal_cleanup);      // we want core dumps
-       // signal(SIGBUS, signal_cleanup);
 
        /*
         * Do not shut down the server on broken pipe signals, otherwise the
@@ -1166,31 +1158,6 @@ int convert_login(char NameToConvert[]) {
 /* 
  * This loop just keeps going and going and going...
  */
-/*
- * FIXME:
- * This current implimentation of worker_thread creates a bottle neck in several situations
- * The first thing to remember is that a single thread can handle more than one connection at a time.
- * More threads mean less memory for the system to run in.
- * So for efficiency we want every thread to be doing something useful or waiting in the main loop for
- * something to happen anywhere.
- * This current implimentation requires worker threads to wait in other locations, after it has
- * been committed to a single connection which is very wasteful.
- * As an extreme case consider this:
- * A slow client connects and this slow client sends only one character each second.
- * With this current implimentation a single worker thread is dispatched to handle that connection
- * until such times as the client timeout expires, an error occurs on the socket or the client
- * completes its transmission.
- * THIS IS VERY BAD since that thread could have handled a read from many more clients in each one
- * second interval between chars.
- *
- * It is my intention to re-write this code and the associated client_getln, client_read functions
- * to allow any thread to read data on behalf of any connection (context).
- * To do this I intend to have this main loop read chars into a buffer stored in the context.
- * Once the correct criteria for a full buffer is met then we will dispatch a thread to 
- * process it.
- * This worker thread loop also needs to be able to handle binary data.
- */
- 
 void *worker_thread(void *arg) {
        int highest;
        CitContext *ptr;