Change stats to let if run even if some of the machines is down.
And also to tolerate temporary outages without keeping
remote machines as permanently down


Notes:
Mon Mar 15 12:50:08 EST 2004 rsc
    It's not clear I got to this patch in time.
    It looks like there is code looking in ether0/0/ifstats
    that has gone away and I can't tell if that's on 
    purpose or not.
    


Reference: /n/sources/patch/sorry/stats
Date: Mon Apr  5 23:23:14 CES 2004
Reviewed-by: rsc

--- /sys/src/cmd/stats.c	Mon Apr  5 23:23:15 2004
+++ /sys/src/cmd/stats.c	Mon Apr  5 23:23:15 2004
@@ -575,7 +575,7 @@
 	return fd;
 }
 
-void
+int
 initmach(Machine *m, char *name)
 {
 	int n, fd;
@@ -583,10 +583,9 @@
 	char *p, mpt[256], buf[256];
 
 	p = strchr(name, '!');
-	if(p){
+	if(p)
 		p++;
-		m->name = estrdup(p+1);
-	}else
+	else
 		p = name;
 	m->name = estrdup(p);
 	m->remote = (strcmp(p, mysysname) != 0);
@@ -597,7 +596,7 @@
 		fd = connectexportfs(name);
 		if(fd < 0){
 			fprint(2, "can't connect to %s: %r\n", name);
-			killall("connect");
+			return 0;
 		}
 		/* BUG? need to use amount() now? */
 		if(mount(fd, -1, mpt, MREPL, "") < 0){
@@ -605,7 +604,7 @@
 			strcpy(mpt, "/n/sid");
 			if(mount(fd, -1, mpt, MREPL, "") < 0){
 				fprint(2, "stats: mount %s on %s failed: %r\n", name, mpt);
-				killall("mount");
+				return 0;
 			}
 		}
 	}
@@ -638,11 +637,6 @@
 
 	snprint(buf, sizeof buf, "%s/net/ether0/ifstats", mpt);
 	m->ifstatsfd = open(buf, OREAD);
-	if(m->ifstatsfd < 0){
-		/* try the old place - this code will disappear on Nov 18th - presotto */
-		snprint(buf, sizeof buf, "%s/net/ether0/0/ifstats", mpt);
-		m->ifstatsfd = open(buf, OREAD);
-	}
 	if(loadbuf(m, &m->ifstatsfd)){
 		/* need to check that this is a wavelan interface */
 		if(strncmp(m->buf, "Signal: ", 8) == 0 && readnums(m, nelem(m->netetherifstats), a, 1))
@@ -661,6 +655,7 @@
 		if(loadbuf(m, &m->bitsybatfd) && readnums(m, 1, a, 0))
 			memmove(m->batterystats, a, sizeof(m->batterystats));
 	}
+	return 1;
 }
 
 jmp_buf catchalarm;
@@ -720,11 +715,19 @@
 			if(display != nil)	/* else we're still initializing */
 				eresized(0);
 		}
-		m->disable = 1;
+		if (m->disable++ >= 5)
+			m->disable = 0; /* give it another chance */
 		memmove(m->devsysstat, m->prevsysstat, sizeof m->devsysstat);
 		memmove(m->netetherstats, m->prevetherstats, sizeof m->netetherstats);
 		return;
 	}
+	snprint(buf, sizeof buf, "%s", m->name);
+	if (strcmp(m->name, buf) != 0){
+		free(m->name);
+		m->name = estrdup(buf);
+		if(display != nil)	/* else we're still initializing */
+			eresized(0);
+	}
 	if(m->remote){
 		notify(alarmed);
 		alarm(5000);
@@ -987,7 +990,7 @@
 	present[which] = 0;
 }
 
-void
+int
 addmachine(char *name)
 {
 	if(ngraph > 0){
@@ -998,8 +1001,11 @@
 		nmach = 0;	/* a little dance to get us started with local machine by default */
 	mach = erealloc(mach, (nmach+1)*sizeof(Machine));
 	memset(mach+nmach, 0, sizeof(Machine));
-	initmach(mach+nmach, name);
-	nmach++;
+	if (initmach(mach+nmach, name)){
+		nmach++;
+		return 1;
+	} else
+		return 0;
 }
 
 void
@@ -1259,9 +1265,9 @@
 		initmach(&mach[0], mysysname);
 		readmach(&mach[0], 1);
 	}else{
-		for(i=0; i<argc; i++){
-			addmachine(argv[i]);
-			readmach(&mach[i], 1);
+		for(i=j=0; i<argc; i++){
+			if (addmachine(argv[i]))
+				readmach(&mach[j++], 1);
 		}
 	}