Cisco:

• Some multicast packets broadcast despite session already being active
  This was recently mentioned in a bugtraq post.

IBM:

• AIX 4.3.3 nfso manpage has a typo
  Manpage specifies nfs_use_reserve_ports, but the actual option is nfs_use_reserved_ports.

Microsoft:

• Excel gets incorrect value and doesn't warn user.
• Exchange server doesn't thread based on In-reply-to: headers.

RedHat Linux:

• anaconda (the installer) dies if /var/mail is partitioned off
  Turns out that /var/mail is supposed to be a symlink to /var/spool/mail. Reported against RH 8.0 in Bug 81127. Fixed by blacklisting /var/mail as a valid partition.
• ntp (port 123/udp) isn't allowed through firewall even though installer says it is (RH 8.0)
• anaconda doesn't support graphical kickstart intsalls over ftp/http
  Reported against RH 8.0 in Bug 87578. They're considering adding this functionality.
• anaconda doesn't follow HTTP redirects to find a kickstart config file
  Reported against RH 9 in Bug 91488.
• /etc/ntp/ntpservers is created, but never referenced
  Reported against RH 9 in Bug 91619.
  Documentation updated.
• dump can't compress to remote tape drive
  Reported against RH 9 in Bug 92049.
• mt retension hangs machine with STT3401A drive
  Reported against RH 9 in Bug 92051.
• Missing file in postgresql-libs
  Reported against RH 9 in Bug 92107
• Flag screensaver hangs X with nVidia drivers
  Reported against RH 8.0 in Bug 91866
• cups has excessive logging
  Reported against RH 9 in Bug 88393
• anaconda ignores -kernel directive
  Reported against RH 9 in Bug 102031
  Fixed in CVS to not unilaterally select "kernel" if we've selected another kernel package.
• logwatch is ignoring date range for some logs
  Reported against RH 9 in Bug 99438
  
• OpenSSH is reporting authentication failures, even though authentication succeeds
  Reported against RH 9 in Bug 101157
  Reported against RH 9 in Bug 102221
  
• /bin/mail crashes on large mailboxes
  Reported against RH 9 (and 7.3) in Bug 103482
  
• LVM snapshots of ext3 filesystems can't be mounted
  Reported against RH 9 in Bug 84278
  Reported against RH 9 in Bug 88115
  Reported against RH 9 in Bug 97843
  
• crontab -l keeps extraneous headers
  Reported against RH 9 in Bug 103899
  Should be fixed in vixie-cron-3.0.1-94
• metaballs dumps core frequently
  Reported against RH 9 in Bug 86054
• rpm hangs with database corruption
  Reported against RH 8.0 and 9 in Bug 73097
• kbackgammon gives sig11
  Reported against RH 9 in Bug 106823
• 1920x1200 not an allowed screen resolution in installer
  Reported against Fedora Core 1 in Bug 108006
• xrandr gives incorrect refresh rate for LCD panels (59Hz)
  Reported against Fedora Core 1 in Bug 108008
• battstat-applet sucks CPU
  Reported against Fedora Core 1 in Bug 104292
• tripwire checks files that don't exist
  Reported in Bug 54867
• ypserv dies randomly
  Reported in Bug 98531
• X dies randomly
  Reported in Bug 90615
• dump provides misleading time estimates
  Reported in Bug 116028
• xfig gives font warnings
  Reported in Bug 116542
• yum package missing from 8.0 legacy-utils
  Reported in Bug 1345
• lvm module of kernel gives an Oops when modifying snapshots
  Reported in Bug 125837
• inadequate logging during TLS sessions
  Reported in Bug 139212
• I've given up recording all of the bugs I find here. It's simply too difficult to maintain this page given how many bugs I find. Instead, you can search for my bugs in RedHat's bugzilla

SGI:

• /etc/init.d/xdm stop will killall itself
  /etc/init.d/xdm stop has the line
  /sbin/killall -k 10 -TERM xdm
  so the script kills itself (since it is named xdm also) and can''t kill a mildly-hung xdm.
  Seen on IRIX 6.2, 6.5.6m, and 6.5.14f.
• syslogd -N doesn't work as advertised under IRIX 6.5.15m
  According to the syslogd manpage, the -N option prevents the host from listening on port 514/udp for syslog messages.
  But it's doing more -- it''s blocking outgoing messages also.
  So that prevents remote loghosting from client machines.
  Opened case number 2330786
  Bug 857459 fixed in 6.5.18
• fsr reports negative file sizes for large (>2GiB) files
  fsr[9627]: insufficient freespace for: ino=92920441: size=-866238919: ignoring Opened case number 233304
  Bug 858139
• fw_teTeX had broken fonts
• system uptime stops incrementing after it's been up a long time (IRIX 6.2)
  Turns out this one had been reported back in '99 (I didn't discover it until 2002.) According to Bug 670377, they use 32 bits for clock_t, so after 2^31 hundredths of a second (248 days) it wraps to negative and stops reporting properly.
  If I remember, I'll take a look at this system after it's been up for 500 days and see if it comes back. ;)
  Machine was decommissioned before 500 days passed
• vi(1) manpage has a see also vi(1).
• SetPerClientControls failed reported in SYSLOG whenever a user logs out
  Bug 853898 fixed in 6.5.18
• User can cause kernel panic in some cases (information withheld for security reasons)
• Kernel tunable parameters tcpiss_md5_secret_lo and _hi cannot be set
  Reported to security-alert@sgi.com
  No response: reported again 6 weeks later.
  Bug 882293 fixed in 6.5.20
  There's also a workaround: edit mtune/bsd and set the -1 to 0

Free Software:

• OpenSSH: Improper handling of command-line flags in ssh-agent.
  From the OpenSSH ChangeLog:
  - markus@cvs.openbsd.org 2002/04/02 11:49:39
    [ssh-agent.c]
    check $SHELL for -k and -d, too;
    http://bugzilla.mindrot.org/show_bug.cgi?id=199
• GCC: "sorry, not implemented: `tree_list' not supported by dump_type" error message in gcc-2.95.
  Bug reported to mailing list Aug 9, bug report filed Aug 11, fixed in 2.95.1 on Aug 18.
• Numerical Python (NumPy): LinearAlgebraError: SVD did not converge
  This can't happen because, by definition, an SVD cannot fail.
  I traced the problem to a max_iteration count that is too small in dlapack_lite.c, line 22562. Increasing the parameter from 20 to 100 fixes the issue.
  Filed bug 601052, and it was fixed in CVS the next day (2002/08/28).
• OpenSSL: make test isn't safe for parallel make
• OpenDX wouldn't compile under RH9: src/exec/libdx/mem.c:382: undefined reference to `errno'
• Perl: make test breaks permissions on /dev/tty
  Reported in bug 24521 on Nov 18, 2003.
  Finally acknowledged and fixed on Feb 29, 2004. Turned out to be a bug in perl itself!
• ClamAV-milter: if max-children is hit clamav-milter <=0.70v will wait for the full 60 seconds even if a child thread is freed up.
  Submitted a patch to the clamav-users mailing list.
  Patch incorporated into 0.70x on May 9, 2004.
• ClamAV-milter: the --local option was for 192.168/16, 10.0.0/24, and 127.0.0.1.
  Submitted a patch to make it for 127.0.0.1, 192.168/16, 172.16/12, and 10/8.
  Also gave a patch to use extended regexps rather than basic, which helps constrain these regexps a little better.
  Patches incorporated into 0.75f on Aug 5, 2004 and 0.75g on Aug 6, 2004.
• Freshclam: doesn't give appropriate warnings when it succeeds on a retry, and doesn't always retry when it should.
  Reported Aug 14.
• ClamAV init scripts start too late in the boot sequence.
  Patches incorporated into CVS Sep 8, 2004.
• SpamAssassin 2.64: sa-learn --forget on an unlearned bounce message causes crash.
  Reported in bug 3819.
• ClamAV JPEG scanning had multiple minor bugs.
  Patches incorporated into CVS Oct 13 and Oct 15, 2004. Additional patches pending.
• LibClamAV gives a warning if it encounters an = in an email header.
• ClamAV-milter: an attacker can abuse the milter to DoS a mailserver with a single connection
  Reported to vendor, and after a new release was made available, reported to Bugtraq
  Currently tracked as CAN-2005-2070
• SpamAssassin 3.0.4: Received parser misparses a Sendmail header (letting spammers bypass SA completely) Reported in bug 4813.

Other:

• Opera gives misleading information when warning of expired certificates. Reported against Opera 7.22 in bug 140453.
• Barracuda Spam Firewall likes to assign random scores to messages at times.