CVS commit: pkgsrc/mail/spamdyke

[Amitai Schlair <schmonz%netbsd.org@localhost>]

Module Name:    pkgsrc
Committed By:   schmonz
Date:           Sat Oct 18 04:51:01 UTC 2008

Modified Files:
        pkgsrc/mail/spamdyke: Makefile distinfo

Log Message:
Update to 4.0.7. From the changelog:

VERSION 4.0.7: 10/17/2008
  Changed Makefile.in to compile configuration.c in two steps: first use gcc
    to produce the preprocessed source, then use gcc to compile it.  For some
    reason, gcc crashes on FreeBSD 6.0 when the file is compiled in one step.
    Thanks to K. Shantanu for reporting this one and Felix Buenemann for
    suggesting the fix.

VERSION 4.0.6: 10/16/2008
  Fixed a problem in examine_ip_in_rdns_keyword_entry() that was not correctly
    terminating the end of the keyword buffer, causing strstr() to search too
    far, leading to false negatives (and potentially segmentation faults).
    Thanks to Erald Troja for reporting this one.
  Fixed another problem in middleman() that was not correctly replacing _all_
    of qmail's AUTH advertisements when the "smtp-auth-level" option is
    "always" or "always-encrypted".  Thanks to Youri Kravatsky for reporting
    this one (again).
  Fixed the fix to a bug in nihdns_query() that was setting
    return_target_name_index to 0 in all cases.  This was causing log messages
    to print the first RBL/RHSBL name instead of the one that actually matched.
    Thanks to Arthur Girardi for reporting this one (again).
  Reverted a change from 4.0.5 -- removing the usable_buf_input flag from
    middleman() meant could only tell if there was input in the buffer, not if
    any of it was actually usable.  If the remote server delays sending its
    data for any reason, middleman() will loop rapidly to continually check if
    its buffered data can be sent to qmail.  Removing the flag meant spamdyke
    was consuming 100% CPU while receiving messages with large attachments.
    Thanks to Paulo Henrique Fonseca for reporting this one.
  Added the "cputime" program to the "tests" folder to measure the CPU time
    used by a process.  Neither the shell "time" command nor the POSIX "time"
    command seem to do that.
  Changed sendrecv to always wait() for its child processes so CPU accounting
    will be performed correctly.
  Fixed check_rhsbl() to correctly return the name of the matching RHSBL instead
    of an index that could be beyond the end of the array.
  Changed the values of LOG_USE_CONFIG_TEST, LOG_USE_STDERR and LOG_USE_SYSLOG
    to make none of them equal to 0.  Because the "log-target" option is a
    CONFIG_TYPE_NAME_MULTIPLE option, it is set to 0 until the command line and
    all configuration files are parsed.  When LOG_USE_CONFIG_TEST is 0, the
    progress messages from process_config_file() are sent to stderr until the
    configuration file is completely loaded.  For Plesk users, xinetd sends
    stderr to the network connection, so the remote server gets the output.
    Thanks to Arthur Girdari for reporting this one and helping track it down.


To generate a diff of this commit:
cvs rdiff -r1.19 -r1.20 pkgsrc/mail/spamdyke/Makefile \
    pkgsrc/mail/spamdyke/distinfo

Please note that diffs are not public domain; they are subject to the
copyright notices on the relevant files.