Re: [Mixmaster-devel] [ mixmaster-Bugs-597688 ] bug in remailer-stats

[Peter Palfrader <peter@pa...>]

On Tue, 20 Aug 2002, noreply@... wrote:

> in remailer-stats replay there is always peek at 0h:
>=20
> Number of messages in the past 24 hours:
>    21h:    Mix:  74     PGP:   76  [Pool size:  26]
>    22h:    Mix:  97     PGP:   29  [Pool size:  26]
>    23h:    Mix:  71     PGP:   32  [Pool size:  26]
>     0h:    Mix:1705     PGP: 1057  [Pool size:  25]
>     1h:    Mix: 161     PGP:  129  [Pool size:  50]
>     2h:    Mix: 100     PGP:   76  [Pool size:  36]
>=20
> I believe this is a bug.

It is.

The problem lies in expiring hourly stats to daily stats: Once a dataset
is older than 24 hours it gets moved to the daily bucket. Unfortunatly
daily buckets are aligned at 00:00GMT each day. So next time stats.log
is read that 00:00GMT sample is within our 24hrs again.

This patch seems to fix the issue. I would apprechiate if someone could
test that it does indeed fix the problem on their remailer too.

The idea is to mark daily sets as such and to never import them to
hourly statistics.

It will take about one days until you get fixed stats.

I am looking forward to success reports.

Index: stats.c
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
RCS file: /cvsroot/mixmaster/Mix/Src/stats.c,v
retrieving revision 1.11
diff -u -r1.11 stats.c
--- stats.c	22 Aug 2002 05:25:22 -0000	1.11
+++ stats.c	22 Aug 2002 17:43:09 -0000
@@ -58,7 +58,8 @@
   long updated =3D 0, havestats =3D 0;
   int msgd[3][24], msg[3][80];
   int poold[2][24], pool[2][80];
-  int i, num, type;
+  int i, num, type, assigned, daysum;
+  char c;
   idlog_t idbuf;
=20
   now =3D (time(NULL) / (60 * 60) + 1) * 60 * 60;
@@ -78,11 +79,14 @@
       case '0':
       case '1':
       case '2':
-	sscanf(line, "%d %d %ld", &type, &num, &then);
-	if (now - then < 0)
+	c =3D '\0';
+	assigned =3D sscanf(line, "%d %d %ld %c", &type, &num, &then, &c);
+	daysum =3D (assigned =3D=3D 4 && c =3D=3D 'd');
+
+	if (now - then < 0 || (daysum && today - then < 0))
 	  break;		/* keep memory consistent even if the time
 				   suddenly goes backwards :) */
-	if (now - then < SECONDSPERDAY)
+	if (now - then < SECONDSPERDAY && !daysum)
 	  msgd[type][(now - then) / (60 * 60)] +=3D num;
 	else if (today - then < 80 * SECONDSPERDAY)
 	  msg[type][(today - then) / SECONDSPERDAY] +=3D num;
@@ -90,10 +94,13 @@
 	  havestats =3D then;
 	break;
       case 'p':
-	sscanf(line, "p %d %d %ld", &num, &i, &then);
-	if (now - then < 0)
+	c =3D '\0';
+	assigend =3D sscanf(line, "p %d %d %ld %c", &num, &i, &then, &c);
+	daysum =3D (assigned =3D=3D 4 && c =3D=3D 'd');
+
+	if (now - then < 0 || (daysum && today - then < 0))
 	  break;
-	if (now - then < SECONDSPERDAY) {
+	if (now - then < SECONDSPERDAY && !daysum) {
 	  poold[0][(now - then) / (60 * 60)] +=3D num;
 	  poold[1][(now - then) / (60 * 60)] +=3D i;
 	} else if (today - then < 80 * SECONDSPERDAY) {
@@ -145,10 +152,10 @@
   for (i =3D 0; i < 80; i++) {
     for (type =3D 0; type < 3; type++)
       if (msg[type][i] > 0)
-	fprintf(s, "%d %d %ld\n", type, msg[type][i],
+	fprintf(s, "%d %d %ld d\n", type, msg[type][i],
 		today - i * 24 * 60 * 60);
     if (pool[0][i] > 0)
-      fprintf(s, "p %d %d %ld\n", pool[0][i], pool[1][i],
+      fprintf(s, "p %d %d %ld d\n", pool[0][i], pool[1][i],
 	      today - i * 24 * 60 * 60);
   }
   unlock(s);

					yours,
					peter

--=20
 PGP signed and encrypted  |  .''`.  ** Debian GNU/Linux **
    messages preferred.    | : :' :      The  universal
                           | `. `'      Operating System
 http://www.palfrader.org/ |   `-    http://www.debian.org/

[Mixmaster-devel] Autoconf and IDEA

[Erik Arneson <erik@aa...>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: RIPEMD160

Peter and I were talking about this, and we aren't sure which behavior
would be best for the configure script.  Here are the options we
thought of:

1. If IDEA support can't be found, configure will exit with an error
   unless --disable-idea is explicitly set.  It will print out a
   little message telling the user to read the docs so they will
   understand what missing IDEA support will do to them and their
   pets.

2. If IDEA support is looked for but not found, spit out a warning and
   continue on.

Which looks better to you guys?  I remember that when I first compiled
Mixmaster I was very frustrated by the IDEA stuff, and I wished
something had pointed out to me what I was missing.  I'm afraid that a
warning might get lost in the configure script's verbosity.

Input would be appreciated!

- -- 
;; Erik Arneson <erik@...>      JD, Ashland Lodge No. 23 ;;
;; GPG Key ID: 2048R/8B4CBC9C    RAM, Siskiyou Chapter No. 21 ;;
;; <http://erik.arneson.org/>;     <http://www.aarg.net/mason>; ;;

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.7 (GNU/Linux)
Comment: Processed by Mailcrypt 3.5.6 and GnuPG

iQEVAwUBPWcRx9tvZZqLTLycAQMY8Af9FpUmiItNRb2xjih1dnyfphlYq+znOASk
4R5KmCFkBf0Mpw218NWw2IpB2hrdKEhbtp5l6L/sqZPquGbGD8B/YzsGw8YpSa+b
RHMfGI6Wy0mqLorCDg2FqRar616ieZZZxon6IyQsskQM+O8oQE4K2XcTGy4SDVxe
ylTgPhSnt6tXS/jcxo5IqQcv8Dfq0QmB2ss+Gm5wi7DNHCcliYRzG3b5LhoQnrm/
XjItyb3Fc+84+rK6nUK5EtJlTpoe79IL8nZdS/5is4rcYMb1XLXcSQn2vl1gFM1p
iegNkzGzGFuevRiI0t7iYNsVAGEnZSJ+ysc2R3YiDcpMWFMDqF1SWA==
=8HG7
-----END PGP SIGNATURE-----