Re: Messages logged by Squid 2.2S4 on Digital UNIX 4.0D

From: Henrik Nordstrom <hno@dont-contact.us>
Date: Sat, 18 Sep 1999 18:09:21 +0200

R Scott Butler wrote:

> Problem/Question: Since the upgrade, the cache.log files have become quite
> large in comparison to their pre-upgrade size as a result of being filled
> with messages like the following:
>
> 1999/09/17 07:42:24| sslWriteServer: FD 26: write failure: (35) Operation
> would block.

Only a logging mishap. These messages should not be logged at loglevel
1. The "error" is perfectly normal, and not an failure in any way.

Attached a small patch. It should apply fine to 2.2.STABLE4.

--
Henrik Nordstrom
Spare time Squid hacker

Sat Sep 18 18:07:48 CEST 1999
Modified Files in squid/src
        ssl.c

Use a lower loglevel for ignored socket write "errors" like EWOULDBLOCK. Loglevel 1
should only log such errors which indeed is errors.

Index: squid/src/ssl.c
diff -u squid/src/ssl.c:1.1.1.20.2.1 squid/src/ssl.c:1.1.1.20.2.2
--- squid/src/ssl.c:1.1.1.20.2.1 Sat Aug 7 22:43:42 1999
+++ squid/src/ssl.c Sat Sep 18 18:07:47 1999
@@ -281,10 +281,14 @@
     }
     cbdataLock(sslState);
     if (len < 0) {
- debug(50, 1) ("sslWriteServer: FD %d: write failure: %s.\n",
- fd, xstrerror());
- if (!ignoreErrno(errno))
+ if (!ignoreErrno(errno)) {
+ debug(50, 1) ("sslWriteServer: FD %d: write failure: %s.\n",
+ fd, xstrerror());
             comm_close(fd);
+ } else {
+ debug(50, 4) ("sslWriteServer: FD %d: write failure: %s.\n",
+ fd, xstrerror());
+ }
     }
     if (cbdataValid(sslState))
         sslSetSelect(sslState);
Received on Sat Sep 18 1999 - 14:51:42 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:48:28 MST