[EMAIL PROTECTED] (Chin Fang) writes:
> In the past two hours, I was reading qmail-popup.c and thus found out
> the underlined line below was doing something unusual:
It is the behavior documented in the man page, though I'm not really
sure why it's written that way.
> Once deleted, I could get *some* :( stuff written by qmail-pop3d to
> STDERR logged via multilog, rather than spitted out to POP clients.
> The reason I said *some* is because I am stuck again. The 2nd
> strerr_warn1 in the attached patch below (WARNING! unfinished) simply
> stays silent. The first one prints its msg correctly. I have been
> looking up and down the src to see what I did wrong, but so far can't
> spot anything (yet).
There's no obvious reason why both calls to strerr_warn1() shouldn't
print, and it's hard to guess without more detailed information. The
relevant portion of a trace along with your run file and the uname -a
of your system would be useful.
[ I wrote ]
> > where stderrto is a small program that opens its first argument for
> > writing on fd 2 and execs the rest of its arguments[1]. In
> > retrospect, patching qmail-popup might have been a better idea.
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
I just tried this, along with my own version of the patch you're
writing, and things works as expected on my machine. (Yes, this is
extremely weak assurance. :)
The diffs to qmail-pop3d.c and qmail-popup.c are appended to this
message.
^L
--
Louis Theran
Nokia Research Center
(Not speaking for Nokia)
Patch to change the way qmail-popup treats stderr:
*** qmail-popup.c.orig Sat May 20 20:03:27 2000
--- qmail-popup.c Sat May 20 20:04:07 2000
***************
*** 88,94 ****
int wstat;
int pi[2];
- if (fd_copy(2,1) == -1) die_pipe();
close(3);
if (pipe(pi) == -1) die_pipe();
if (pi[0] != 3) die_pipe();
--- 88,93 ----
Patch to make qmail-pop3d log the number of bytes slurped in RETR
commands:
*** qmail-pop3d.c.orig Sat May 20 19:54:14 2000
--- qmail-pop3d.c Sat May 20 20:00:50 2000
***************
*** 17,22 ****
--- 17,37 ----
#include "timeoutread.h"
#include "timeoutwrite.h"
+ #include "strerr.h"
+
+ /* @@@ patched to log the number of bytes per connection.
+ *
+ * A simple patch to make qmail-pop3d emit log lines of the form:
+ *
+ * qmail-pop3d <pid>: <nbytes> message bytes read
+ *
+ * to stderr, which is assumed to be pointed at multilog
+ *
+ * DON'T BLINDLY ASSUME THIS WORKS, it has undergone only cursory testing
+ *
+ * --Louis Theran, 2000-05-20
+ */
+
void die() { _exit(0); }
int saferead(fd,buf,len) int fd; char *buf; int len;
***************
*** 35,40 ****
--- 50,79 ----
return r;
}
+ /* @@@ keep track of how many bytes we slurp from messages in pop3_top() */
+ static unsigned long msgbytesread = 0;
+
+ static int counted_read(fd,buf,len) int fd; char *buf; int len;
+ {
+ int rv;
+
+ rv = read(fd,buf,len);
+ if (rv > 0) msgbytesread += rv;
+ return rv;
+ }
+
+ /* @@@ generating log messages */
+ char msgbytes_str[FMT_ULONG];
+ char pid_str[FMT_ULONG];
+
+ static void log_bytes()
+ {
+ msgbytes_str[fmt_ulong(msgbytes_str,msgbytesread)] = 0;
+ pid_str[fmt_uint(pid_str,getpid())] = 0;
+
+ strerr_warn5("qmail-pop3d ",pid_str,": ",msgbytes_str," message bytes read",0);
+ }
+
char ssoutbuf[1024];
substdio ssout = SUBSTDIO_FDBUF(safewrite,1,ssoutbuf,sizeof ssoutbuf);
***************
*** 193,198 ****
--- 232,238 ----
rename(m[i].fn,line.s); /* if it fails, bummer */
}
okay();
+ log_bytes();
die();
}
***************
*** 268,274 ****
fd = open_read(m[i].fn);
if (fd == -1) { err_nosuch(); return; }
okay();
! substdio_fdbuf(&ssmsg,read,fd,ssmsgbuf,sizeof(ssmsgbuf));
blast(&ssmsg,limit);
close(fd);
}
--- 308,314 ----
fd = open_read(m[i].fn);
if (fd == -1) { err_nosuch(); return; }
okay();
! substdio_fdbuf(&ssmsg,counted_read,fd,ssmsgbuf,sizeof(ssmsgbuf));
blast(&ssmsg,limit);
close(fd);
}
