Actually this is more of a QIFAM (Questions I Frequently Ask
  Myself)... various tidbits of mostly computer-related info, that for
  one reason or another I tend to forget. Usually that's because it's
  information that I use infrequently enough that it never gets
  imprinted in my meory, so rather than have to look it up from first
  sources each time I do need it, I just write it down here.  

  for f in "$@"; do
      a2ps --landscape $f --columns=1 --chars-per-line=120 --output=${f%.*}.ps
  done  

• parameter substitution

   if the invocation is

      ./somescript "1 2 3" a b c

   then

        $#   is  4
        $*   has 6 parameters
       "$*"  has 1 parameter
        $@   has 6 parameters
       "$@"  has 4 parameters

   so, for forwarding parameters from one script to another, the
   construct to use is "$@", including the double-quotes    

• pattern-matching

   shopt -s nullglob  # patterns with no matches expand to the empty string
   shopt -u nullglob  # patterns with no matches stay unchanged    

• readline history manipulation

   Esc-.   retrieves the last argument from the previous command;
           subsequent presses retrieve the last argument from commands
           further back in the history    

• variable manipulation
• left-stripping:

    ${var#pattern}  remove the shortest matching piece
    ${var##pattern} remove the longest matching piece      

• right-stripping:

    ${var%pattern}  remove the shortest matching piece
    ${var%%pattern} remove the longest matching piece      

• example:

    var=/foo/bar/baz.tar.gz

    ${var%.*}    --> /foo/bar/baz.tar
    ${var%%.*}   --> /foo/bar/baz
    ${var#*/}    -->  foo/bar/baz.tar.gz
    ${var##*/}   -->          baz.tar.gz      

• substitution:

    ${var/pattern/string}   replace first 'pattern' match with 'string'
    ${var//pattern/string}  replace all 'pattern' matches with 'string'      

• line-oriented reading

   read directly from a file into a file descriptor:

   exec 6<fname
   while read -u 6 line; do
      echo "$line"
   done

   pipe a subshell's output into line-oriented reading:

   (
       svn status
   ) | (
       while read line; do
           case "$line" in
               M*)
                   echo $line
                   ;;
           esac
       done
   )    

  P(A,B) = P(A|B) * P(B)
	 = P(B|A) * P(A)

  => P(A|B) = P(B|A) * P(A) / P(B)

     posterior = likelihood * prior / constant  

• DEBUGGING things to try when debugging
• check lifetimes of objects referred to by local references and pointers
• use Electric Fence

    Just link to libefence (i.e. add "-lefence" to the link
    command-line).  Its best minimize the size of the input to the
    program, since Electric Fence is a huge memory hog.

    You can also use dynamic linking. If you're using a Bourne shell,
    the statement export LD_PRELOAD=libefence.so.0.0 will cause
    Electric Fence to be loaded to run all dynamic executables.  The
    command 'ef command' runs a single command under Electric Fence.      

• use libgmalloc under Mac OSX (try "man libgmalloc")
• use the MALLOC_CHECK_ environment variable with glibc

    export MALLOC_CHECK_=1   # print info to stderr about bad malloc/free
    export MALLOC_CHECK_=2   # abort immediately if bad malloc/free
    export MALLOC_CHECK_=3   # combination of (1)+(2)      

• setjmp/longjmp

#include <setjmp.h>
#include <cstdio>

int main()
{
  jmp_buf state;

  int val = setjmp(state);
  if (val == 0)
    {
      printf("direct return\n");
    }
  else
    {
      printf("indirect return %d\n", val);
      return 0;
    }

  longjmp(state, 3);
}    

• POINTERS pointers to constants and constant pointers:

char *const cp;                 // const pointer to char
char const* pc;                 // pointer to const char
const char* pc2;                // pointer to const char

read the declaration left to right    

  getting started:

  need to edit /etc/omniORB.cfg to include the following line:

    InitRef = NameService=corbaname::127.0.0.1

  need this in ~/.bashrc so omniNames can run as non-root:

    export OMNINAMES_LOGDIR=$HOME/local/var/omninames

  then run omniNames as non-root with

    omniNames -start # for the first time
    omniNames        # for subsequent times  

to reload crontab file       crontab ~/.crontab  

• the modules file

   Here's what I've learned about the modules file, and in particular
   ampersand modules.

   I had the following file directories in my $CVSROOT:

     $CVSROOT/pkgs/mx/
     $CVSROOT/pkgs/optim/

   containing sets of C++ source files that I wanted to be able to
   reuse in different projects, via different modules.

   As well as

     $CVSROOT/devscripts/

   containing a set of scripts that I also wanted to reuse.

   I wanted to set up an optimization module:

     $CVSROOT/optimization/

   for which the module, when checked out, would also contain the
   following directory structure:

     optimization/devscripts/
     optimization/src/mx/
     optimization/src/optim/

   Here was my first attempt at a modules file:

     pkgs_mx          -d mx     pkgs/mx
     pkgs_optim       -d optim  pkgs/optim

     optimization_src -d src    &pkgs_mx &pkgs_optim

     optimization               optimization &optimization_src &devscripts

   OK, so now I do a 'cvs co optimization'. I get the desired
   directory structure. But then when I 'cd optimization' and 'cvs
   update', cvs appears to not recognize the 'src' or 'devscripts'
   direcgtories. The 'optimization/CVS/Entries' contains only a single
   line:

     D

   I knew that I had seen something like this approach work before in
   my previous efforts with cvs, so on a hunch I did a 'cvs add; cvs
   commit' for a dummy file in the 'optimization' directory (which had
   been empty until now... that is, $CVSROOT/optimization/ was an
   empty directory). Now, when I do a fresh 'cvs co optimization', the
   optimization/CVS/Entries file is still initially devoid of lines
   referring to either devscripts or src... however, if I 'cd
   optimization' and 'cvs update', then afterwards there is a new line
   in CVS/Entries referring to devscripts. But still nothing referring
   to 'src'.

   Exploring more, I peeked inside the
   'optimization/src/CVS/Repository' file and found
   'CVSROOT/EmptyDir'... OK, so cvs is treating the 'optimization_src'
   module specially because it consists only of ampersand directives.

   So, on another hunch, I tried adding a new, empty directory in the
   cvs repository:

     $CVSROOT/dummydir

   and then changed the optimization_src line in the modules file from
   this:

     optimization_src -d src    &pkgs_mx &pkgs_optim

   to this:

     optimization_src -d src    dummydir &pkgs_mx &pkgs_optim

   Now, once more I do a fresh 'cvs co optimization'. As before, there
   are initially no lines in CVS/Entries referring to devscripts or
   src, but now if I 'cd optimization' and 'cvs update', both
   'devscripts' and 'src' are recognized, and I subsequently get lines
   for both in CVS/Entries.

   Lesson:

     Don't compose modules out of ampersand references alone; always
     have at least a real "dummydir" at the beginning... not much of
     a problem since a single "dummydir" will work for the entire
     repository.    

• cheatsheet for rpm/yum <-> apt/dpkg

   rpm -qa              dpkg-query -l
   rpm -ql <pkg>        dpkg-query -L <pkg>
   rpm -qf <file>       dpkg-query -S <file|pattern>
   yum remove <pkg>     apt-get remove <pkg>    

• commands

   forward-char                 C-f
   backward-char                C-b
   previous-line                C-p
   next-line                    C-n

   forward-word                 M-f
   backward-word                M-b
   tab-to-tab-stop              M-i

   next-page                    C-v
   previous-page                M-v

   delete-character             C-d
   kill-word                    M-d
   backword-kill-word           M-DEL

   transpose letters            C-t
   transpose words              M-t
   transpose lines              C-x C-t

   quote message in reply       C-c C-y

   read-only toggle             C-x C-q

   generate etags table         etags *\.[ch]*
   load etags table             M-x visit-tags-table RETURN
   find-tag                     M-.
   tags-loop-continue           M-,
   tags-query-replace           M-x tags-query-replace RETURN
   list-tags                    M-x list-tags RETURN

   beginning-of-defun           M-C-a
   end-of-defun                 M-C-e
   indent-region                M-C-\    

• local variables

   Note that the local variable list must be within 3000 characters of
   the end of the file.

   /* For emacs customization... */
   /* Local Variables: */
   /* tab-width: 5 */
   /* End: */

   The "/*" and "*/" can be arbitrary strings, as long as they are the
   same on all of the lines.    

• markers

   to create: command-option click in lower third of ruler bar
   to delete: grab marker text with cursor and drag downward    

• automation

   to create a curve segment: control-option-drag on the middle of the
   curve (drag left-right or up-down)    

  switch with

  system-config-network-cmd --profile <profilename> --activate

  trash /var/lib/dhclient/dhclient-eth0.leases  

• how to turn off http 'referer' (sic)

   Go to about:config in your address bar and search for this:

     network.http.sendRefererHeader

   and set the value to 0.

   Some unexpected pages might start breaking, so beware.

   http://www.mozilla.org/quality/networking/docs/netprefs.html    

  public class HelloWorld {
      public static void main(String[] args) {
  	System.out.println("Hello World!");
      }
  }

  To make an executable main program:

    gcj HelloWorld.java --main=HelloWorld

  To make a *.class file:

    gcj HelloWorld.java -C

  To make a *.o object file:

    gcj -c HelloWorld.java

  To link C++ and Java code:

    gcj test.cc HelloWorld.o

  To make a CNI C++ header:

    gcjh HelloWorld

  To make a JNI C header:

    gcjh -jni HelloWorld

  To interpret Java bytecode:

    gij HelloWorld  

connect.newtella.net:6346
gnutellahosts.com:6346
router.limewire.com:6346
gnutella.hostscache.com:6346  

• key handling

   gpg --gen-key
   gpg --list-keys

   gpg --export > somekeyfile
   gpg --import somekeyfile    

• basic encrypt/decrypt

   gpg (--recipient|-r) me@here.com (--encrypt|-e) somefile
   gpg (--decrypt|-d) somefile.gpg > somefile

   use (--armor|-a) to make the output be ascii-safe

   use stdin/stdout:

   cat somefile | gpg -r me@here.com -e > somefile.gpg    

• with signature

   gpg --sign --encrypt test.txt
   gpg -o test.txt --decrypt test.txt.gpg    

• to force a particular rank distance between nodes

   use the "minlen" attribute on the joining edge, e.g. minlen=2 to
   separate the nodes by at least two rank levels    

hfspax -wv -f thisfolder.pax -x cpio thisfolder
hfspax -rv -f thisfolder.pax  

  when printer goes buggy, try re-running "HP All-In-One Setup Assistant"  

• Error "Offending Operator: 'ter'" or "Offending Operator: 'und'" occurs when you open an Illustrator document (10.x)

Issue

When you try to open an Adobe Illustrator document, Illustrator
returns one of the following errors:

-- "Can't open the illustration. The illustration contains an illegal
or misplaced operator. Offending Operator: 'ter' Context: :/Dict;
/Part,/Part, /CompoundFilter : (Chain Style Filter) 0 0 /Filter." (for
a CMYK file)

-- "Offending operator: 'und' Context: ; /Dict ; /Part , /BasicFilter
: (Adobe Round Corners) 1 0 /Filter , (Round Corners) /PluginFileName
,(Round)" (for an RGB file)

After you click OK in the error dialog box, the document is blank or
severely corrupted.

Detail
The document contains the Wheelchair Access symbol from the Symbols
palette.

Solutions

Do one of the following:

Solution 1: Place the document into a new Illustrator file.  Create a
new Illustrator file, and then place the damaged document into
it. Make sure that All Documents is selected in the Place dialog box.

Note: If you complete this solution, you will lose all styles,
symbols, brushes, blends, and warps. Other effects will be expanded,
and all layers will be merged.

Solution 2: Recover a plain-text version of the document.  If the file
was saved with compression (the default), you must first recover a
plain-text version of the file, which you can then open in a text
editor.

Note: If you complete this solution, all of the editing capabilities
and layers that were in original file will be retained.

To recover a plain-text version of the document:

1. Exit from Illustrator, and then open the Adobe Illustrator
Preferences file in a text editor (for Windows, this is "Documents and
Settings/<username>/Application Data/Adobe/Adobe Illustrator CS
Settings/AIPrefs").

2. Look for the line "/enableContentRecovery 0" and change the value
from "0" to "1" so that the line reads:

/enableContentRecovery 1

[otherwise try EnableContentRecovery (i.e. with capital 'E')]

3. Start Illustrator, and choose File > Open.

4. Select the damaged document, and then click OK while holding down
Ctrl + Shift + Alt (Windows) or Command + Shift + Option (Mac OS). An
empty document opens. Close the document without saving
it. Illustrator extracts the data from the damaged document and
creates a new, uncompressed, document that has the same name as the
original (preceded by an underscore). Illustrator saves the new
document in the same folder as the original.

5. In a text editor that can save in text-only format (for example,
WordPad) open the new document.

6. Search the document for "Wheelchair Access." Delete the line
preceding the first occurrence (that is, %AI10_BeginSymbol) and all
other lines down to and including the line "%AI10_EndSymbol." Delete
the second occurrence of "Wheelchair Access."

7. Save the document with a new name, and then open it in Illustrator.

Background information

The Wheelchair Access symbol from the default Symbols palette is
damaged; Illustrator can't open files that contain this symbol.

To prevent this problem from recurring, delete the Wheelchair Access
symbol from both of the Adobe Illustrator Startup documents (in the
Plug-ins folder) and from the Symbols palette before you work on a
document.

i.e.:

/Applications/Adobe Illustrator/Plug-ins/Adobe Illustrator Startup_CMYK

and

/Applications/Adobe Illustrator/Plug-ins/Adobe Illustrator Startup_RGB

The line "enableContentRecovery 1" enables Illustrator to decompress
files that it can't open otherwise.    

• animated gif images with "gifsicle"

   http://www.lcdf.org/gifsicle    

• generating postscript from bitmap images

     pnmtops
       [-scale s]
       [-dpi n] (default is 300dpi)
       [-imagewidth n]
       [-imageheight n]
       [-width=N]
       [-height=N]
       [-equalpixels]
       [-turn|-noturn]
       [-rle|-runlength]
       [-nocenter]
       [-nosetpage]
       [pnmfile]

     to make nice eps images e.g. for LaTeX:

       anytopnm img.xxx | pnmtops -nocenter -noturn -equalpixels -nosetpage > img.ps
       eps2eps img.ps img.eps

     Using 'eps2eps' to transform the output of pnmtops helps reduce
     the size of the postscript file since it encodes the bitmap data
     in a more compact format.    

• printing jpeg/jpg digital photos

   jpegtopnm $f > {$f:r}.ppm
   pnmtops -height 5 -width 7 ${f:r}.ppm > ${f:r}.ps
   lpr ${f:r}.ps    

  video
  compression: H.264
  quality: High
  framerate: 15 (or "current")
  bitrate: 1500 kbits/sec
  frame reordering: yes
  encoding mode: multi-pass
  dimensions: 640x480

  audio
  format: AAC
  sample rate: 48.000 kHz
  channels: stero
  bit rate: 256 kbps  

  To set keyboard shortcuts for minimizing windows, etc., go to the
  KDE Control Center, then "Regional & Accessibility", then "Keyboard
  Shortcuts"  

  New Shell        Alt+Ctrl+N
  Switch session   Shift+Left, Shift+Right
  Move session     Shift+Ctrl+Left, Shift+Ctrl+Right

  Change shortcuts Settings->Configure Notifications...  

• apm/acpi

   To use APM instead of ACPI:

   The simplest way to do this is to add the "acpi=off" option to the
   end of the "kernel" line in /boot/grub/grub.conf.

   kernel /boot/vmlinuz-2.6.xxx ro root=LABEL=/ rhgb quiet acpi=off

   (from http://www.cs.umd.edu/~nakamura/misc/t40.html)    

• flash memory

   cat /proc/bus/usb/devices
   cat /proc/scsi/usb-storage/2
   sudo fdisk -l


   # fstab:
   /dev/sda1       /mnt/usb        auto  noauto,user,owner,sync  0 0

   dmesg | grep -i usb    

• kernel modules

   load with /sbin/insmod

   put insmod commands in /etc/rc.d/rc.local to have them loaded at
   boot time

   list installed+loaded modules with /sbin/lsmod    

• kernel building
• unpack sources
• specify EXTRAVERSION at top of top-level Makefile
• specify CC (esp. if need gcc296) in top-level Makefile
• make mrproper
• make clean
• make xconfig
• make dep
• make
• make bzImage
• make modules
• sudo mkdir /lib/modules/*
• sudo make install (BUT: see note below about bogus "initrd" files)
• sudo make modules_install
• configure lilo or grub

    then run "sudo lilo -v -t" to test that lilo will succeed
    then run "sudo lilo -v"    to actually run lilo and update the boot sector      

• notes

    18-Mar-2003 2.4.18-27-rjp failed after the first build+reboot,
                because something was wrong with the initrd file. But,
                just redoing "sudo make install" changed the initrd
                size to normal, and then reboot worked fine. Perhaps
                this is because I had edited lilo.conf in the meantime
                to point to the new vmlinuz file?

    30-Jul-2003 Identical experience to above, this time with
                RH-2.4.20-19.8. (1) Did normal build+install. During
                install, got the usual "no suitable template" error
                from lilo/grubby. Suitably edited lilo.conf by
                hand. (2) Reboot, and get an error just before the
                init sequence would begin, including something like
                "couldn't mount ext3 filesystem" and "try passing an
                init=" option at the command line. (3) Reboot back
                into working kernel. Looked at the initrd for the new
                build, and as before, it was "too small" as compared
                to the other initrd files laying around. (4) Went back
                to kernel build directory and re-ran "sudo make
                install". Now, the initrd file is twice as big. (5)
                Once more reboot into the newly-built
                kernel. Everything works fine now.      

• linux wireless networking
• have to enable CONFIG_FILTER in kernel
• 'lspci' to see which cards were detected
• 'iwconfig' to see wireless interfaces
• 'cardmgr/cardctl' to manage cards
• '/sbin/ifup eth1' and '/sbin/ifdown eth1' to start/stop interfaces
• config files are at /etc/sysconfig/network-scripts/ifcfg-eth[01]
• boot-time startup is controlled by "ONBOOT=" line in ifcfg file
• list open files

   /usr/sbin/lsof    

• list open sockets

   /usr/bin/lsof -VRi

   -i to list internet/network files
   -R to list PPID
   -V be verbose if there were no matches for a given request

   optionally:

   -l to inhibit user-id --> username conversion
   -n to inhibit ip-address --> hostname conversion
   -P to inhibit port-number --> portname conversion    

• disconnect+reconnect process controlling terminals

   start the process with

     dislocate  program_name args...

   then use the escape sequence Ctrl-] to get back into 'dislocate' to have
   the opportunity to disconnect the process from the terminal

   then at a later time, on a different terminal, just do 'dislocate' with
   no args to reconnect to the detached process    

  dd if=/dev/zero of=plaintext.img bs=1M count=20
  losetup /dev/loop0 plaintext.img
  mkfs -t ext2 /dev/loop0
  mount /dev/loop0 mountpoint

  dd if=/dev/urandom of=ciphertext.img bs=1M count=20
  losetup -e aes /dev/loop0 ciphertext.img
  mkfs -t ext2 /dev/loop0
  mount -o loop,encryption=aes ciphertext.img /path/to/mountpoint

  umount /path/to/mountpoint
  losetup -d /dev/loop0

  http://www.linuxjournal.com/article/8599  

• console mode (i.e., no Aqua)

   At the Aqua login window, enter ">console" for username (note the
   first character is a 'greater-than' character)

   This brings up a standard console login screen... enter
   username+password there.

   To run an X session, just run "startx"

   NOTE: after logging out of X session, the system will appear to
   hang... it isn't really, but the console video is confused. Two
   options:

     (1) when the system seems to be hanging, just blind-type
         "logout<return>" and this should bring back the Aqua login
         window.

     (2) instead of starting X with "startx", do "exec startx"
         instead; then when X session quits, the Aqua login will
         return automatically    

• Finder

   to reset default size/view/etc. of the Finder window for a given
   folder, open the folder with Cmd-doubleclick and then change the
   settings as desired    

• gcc build tips for Mac OS X

   BIG KEY: --> unlimit stacksize

   make CC='cc -no-pre-comp' bootstrap    

• kernel panics

   how to get a kernel panic log:

   open /Applications/Utilities/Console application, hit the "Logs" icon in
   the toolbar, open /Library/Logs, and the contents of panic.log    

• keyboard shortcuts

from http://www.uark.edu/compserv/softsys/macintosh/macos/osx/osxkeybd.html

Option-Command-esc      Force Quit
Control-Eject   Restart, Sleep, Shutdown dialog box
Control-Command-Eject   Quit all applications and restart
Option-Command-Eject or
Option-Command-Power    Sleep
Press X during startup          Force Mac OS X startup
Press Option-Command-Shift-Delete
during startup  Bypass primary startup volume and seek a different startup volume (such as a CD or external disk)
Press C during startup  Start up from a CD that has a system folder
Press N during startup  Attempt to start up from a compatible network server (NetBoot)
Press R during startup  Force PowerBook screen reset
Press T during startup  Start up in FireWire Target Disk mode
Press Shift during startup      start up in Safe Boot mode and temporarily disable login items and non-essential kernel extension files (Mac OS X 10.2 and later)
Press Command-V during startup  Start up in Verbose mode.
Press Command-S during startup  Start up in Single-User mode (command line)    

• ldd

   from http://developer.apple.com/technotes/tn2002/tn2071.html:

	The ldd command is not available in Mac OS X; however, you can
	use the command otool -L to get the same functionality that
	ldd does. The otool command displays specified parts of object
	files or libraries. The option -L displays the name and
	version numbers of the shared libraries that the object file
	uses. To see all the existing options, see the manual page for
	otool.    

• printer sharing in Leopard (10.5)

   from: http://docs.info.apple.com/article.html?artnum=306984

   cupsctl BrowseProtocols='"cups dnssd"'    

• shm

   sysctl can't modify shm settings once they have been set the first
   time during boot; therefore, they must be set in /etc/rc with a
   line like this:

   sysctl -w kern.sysv.shmmax=134217728 kern.sysv.shmmin=1 \
	  kern.sysv.shmmni=64 kern.sysv.shmseg=64 kern.sysv.shmall=32768    

• Terminal application

   defaults write com.apple.terminal TerminalOpaqueness '0.75'

   to copy escape sequences into the "Terminal Inspector", do the
   following to get the sequence into the pasteboard:

   echo -ne "\e[5C"|pbcopy    

• startup display resolution problems:

   remove /Library/Preferences/com.apple.windowserver.plist
   remove ~/Library/Preference/ByHost/com.apple.windowserver.*.plist    

• to make an app run with no menubar and no dock icon

   add the following line to Whatever.app/Contents/Info.plist:

   <key>NSUIElement</key><string>1</string>    

• verbose startup

   hold cmd-v (openapple-v) during startup, or, to make verbose startup the
   permanent selction, do:

     sudo nvram boot-args="-v"

     http://www.icsi.berkeley.edu/~wooters/mactips.html

   Or, according to http://mfdh.ca/apple/mac_os_x.html, add lines like the
   following to /etc/rc to make the change really permanent:


      # Setting verbose option for next reboot.
      #
      nvram boot-args="-v"
      ConsoleMessage "Verbose booting engaged."    

  from http://eleccomm.ieee.org/demo1/userdoc.shtml

  A user can subscribe to a list, or ask about his or other users'
  subscriptions, by sending mail to the majordomo server, usually
  majordomo@hostname. Put the commands in the body of the mail message
  (not on the "Subject:" header component). Other than white space,
  the commands must be the first text in the message body; in other
  words, don't begin with "Dear Majordomo."  

• help

   Sends one-line summaries of majordomo commands. The first line of
   output tells you which version of Majordomo the server is running.    

• info list

   Sends an introduction to the list list.    

• subscribe list [address]

   This command subscribes a user to the named list. Unless the user
   includes the optional address, Majordomo will use the e-mail
   address in the mail message header ("Reply-To:", if any, otherwise
   "From:") to send mail to the user.    

• unsubscribe list [address]

   This unsubscribes the user (or address) from list.    

• which [address]

   Tells the lists to which the user (or address) is subscribed.    

• who list

   Lists the subscribers of list.    

• index list

   Gives a listing of files in the list archive. If the list is
   private, only members can run this command.    

• get list filename

   Mails a file named filename from the list archive. If the list is
   private, only members can run this command.    

• end

   Stops reading the message. This is useful for users whose mailer
   adds text (like a signature) to the end of a message. A line that
   starts with a dash (-) is also treated as an end command by newer
   versions of Majordomo. Because many peoples' mail message
   signatures start with some dashes, this avoids error messages.    

warn about undefined variables   --warn

`$@' The file name of the target.

`$<' The name of the first prerequisite.

`$^'
`$+'
     The names of all the prerequisites, with spaces between them.
     For prerequisites which are archive members, only the member
     named is used (*note Archives::).  The value of `$^' omits
     duplicate prerequisites, while `$+' retains them and preserves
     their order.


.PRECIOUS: %.o  

• delayed initialization trick

   REAL_SOURCES = $(shell echo a b c | tee -a log)

   SOURCES_INIT = $(eval SOURCES := $(REAL_SOURCES)) $(SOURCES)

   SOURCES = $(call SOURCES_INIT)

   all: foo1 foo2

   foo1: $(SOURCES)
   	echo foo1: $+

   foo2: $(SOURCES)
   	echo foo2: $+    

• Gaussian gaussian normal distributions

half-width at half-height == sqrt(log(4) * stddev)) ~= 1.177 * stddev
full-width at half-height                           ~= 2.355 * stddev

g = (1 / (sqrt(2*pi) * sigma)) * exp(-0.5 * ((x - mu)/sigma)^2)    

• logarithms

log_a(x) := ln(x) / ln(a)    

• to run matlab in debugger

   "matlab -Dgdb", where "gdb" is the name of the debugger    

• to draw a circle

   h = rectangle;
   get(h);
   set(h, 'Curvature', [1 1]);    

• things that bug me about matlab
• non-uniform grammar

    There isn't the same flexible concept of 'expression' as there is
    in C/C++. In C/C++, anything you can do with a variable you can
    also do with a value returned from a function, or with the result
    of a math operator. But not so in matlab. E.g. if you want to
    return the elements of the variable 'x' rearranged in a single
    column, you can do 'x(:)'. But you can't do the same thing to a
    function result, for example -- if you have a function 'foo' you
    can't do 'foo(3)(:)'.

    To me, this gives the language an uncomfortable sense of
    incompleteness and non-uniformity.      

• to avoid lines, plots, etc. from being clipped outside the plot box

   h=line([x1 x2], [y1 y2]);
   set(h, 'clipping', 'off');    

• Ctrl-u to enter edit mode using keyboard

• envelopes in word

   Page Setup: top faces right
   Envelope: Custom Page Options: "Face up" "top facing left, centered"
   Printer: load envelope with top facing left (go figure...)    

  use munpack/mpack from ftp://ftp.andrew.cmu.edu/pub/mpack/  

• to extract a .wav file from a .mp3

   mpg123 -w myfile.wav myfile.mp3    

• to extract a section of a .wav file

   wavpipe -i foo.wav -t 1:12-2:45 foo.wav    

• to fade in/out a .wav file (destructive, in-place edit)

   wavfade -d -T 1.5 -L 0 -t 1.5 -l 0 foo.wav    

• to mp3-encode a .wav file

   lame -h -b 160 foo.wav foo.mp3    

• to render a figure in a pixel-precise way

   set(gcf, 'units', 'pix', 'position',[100 100 w h]);
   set(gcf, 'PaperPositionMode', 'manual');
   set(gcf, 'PaperPosition', [0 0 w h]);
   print(gcf, '-dpng', '-r1', 'test.png');    

• running a test system as an unprivileged user

/usr/bin/mysql_install_db --datadir=/tmp/mysql
/usr/libexec/mysqld --socket=/tmp/mysql.sock --datadir=/tmp/mysql &
/usr/bin/mysqladmin -u root --socket=/tmp/mysql.sock password 'XXXXXX'
/usr/bin/mysql --socket=/tmp/mysql.sock --user root --password
/usr/bin/mysqladmin -u root -pXXXXXX --socket=/tmp/mysql.sock shutdown    

  on linux:

  hwclock
  service ntpd stop && ntpdate time.nist.gov && service ntpd restart
  /etc/ntp.conf
  service ntpd restart
  ntpq -p; # query peers
  ntpq -p time.nist.gov; ntpq -p clock.redhat.com  

  check for latest driver:

    http://www.nvidia.com/object/unix.html

  extraact the contents of the package:

    sh NVIDIA-Linux-x86-1.0-7676-pkg1.run -x

  probably need to do 'unset ARCH' before running nvidia-installer
  (otherwise it will think the kernel arch is supposed to be i686
  instead of i386)

  to build+install for a currently RUNNING kernel version:

    ./nvidia-installer --kernel-name=[KERNELNAME]

  to build+install for a currently NON-RUNNING kernel version:

    ./nvidia-installer -K --kernel-name=[KERNELNAME]

  where KERNELNAME is what the output of `uname -r` would be if the
  desired kernel were actually running.

  might need to do this:

    cp -a /dev/nvidia* /etc/udev/devices
    chown root.root /etc/udev/devices/nvidia*

  AFTER installing the nvidia driver, but BEFORE
  rebooting... otherwise the boot process might hang at "Configuring
  kernel parameters"

  if glx crashes, check whether /usr/lib/libGLcore.so.N.N.NNNN has
  been modified relative to path/to/NVIDIA-Linux/usr/lib/libGLcore --
  sometimes the file gets corrupted -- in that case, try reinstalling  

   /usr/X11R6/bin/glxinfo    command to report glx extension info

   from Matrox readme: DRI requires modules "agpgart" and "mga" add
   "/sbin/insmod agpgart" and "/sbin/insmod mga" to /etc/rc.local

   Strange crashes occur unpredictably -- terminal shows "Segmentation
   Fault", but no core dump is produced. This seems to be alleviated
   by requesting an indirect rendering context, and possibly by just
   rebooting the machine... maybe this is a problem with the gfx card
   driver/kernel module that handles the direct rendering?  

  collect profiling data:

  sudo opcontrol --reset # clear out data from previous session
  sudo opcontrol --start # start the oprofile daemon (oprofiled)
  ... run /some/long/program
  sudo opcontrol --shutdown # stop profiling

  generate profiling report:

  opreport --demangle=smart --symbols /some/long/program
  opreport --exclude-dependent --demangle=smart --symbols /some/long/program  

  If Parallels NAT and Parallels Host-Guest show self-assigned IP
  addresses, then try the following:

  sudo killall -HUP pvsnatd
  sudo /Library/StartupItems/Parallels/pvsnatd  

• network byte order htonl ntohl

   there are no htonl() or ntohl() functions in perl, but any network
   read/write calls generally involve a perl pack() or unpack()
   followed by a print, and pack() and unpack() have 'n' and 'N'
   templates which can be used to convert a number to a network byte
   order byte string (with pack()) or convert a network byte order
   byte string to a number (with unpack())    

• edit file(s) in place

   $ perl -p -i.bak -e "s/foo/bar/; ..."    

• line-oriented file reading/writing

  #!/usr/bin/perl -w

  open(FOO, "foo") or die "can't open foo for reading: $!\n";

  while ($line = <FOO>) {
      ($f1, $f2) = split(" ", $line);
  }

  open(BAR, ">bar") or die "can't open bar for writing: $!\n";
  # or open(BAR, "<filename"); # same as just "filename"
  # or open(BAR, ">filename");
  # or open(BAR, ">>filename");
  # or open(BAR, "| output-pipe-command");
  # or open(BAR, "input-pipe-command |");    

• running on Mac OSX
• info at http://www.reitter-it-media.de/software/osxpostfix.html

    also http://www.stepwise.com/Articles/Workbench/eart.index.html
    also http://www.cutedgesystems.com/software/PostfixEnabler/

    edit /etc/postfix/master.cf, uncomment the 'smtp' line

    edit /etc/hostconfig, change MAILSERVER from -NO- to -YES-

    start postfix with

      # /System/Library/StartupItems/Postfix/Postfix start
      (as root)

    logs in /var/log/mail.log

    edit /etc/postfix/mail.cf to set myhostname, mydomain, myorigin

    edit /etc/postfix/canonical to set up address rewriting

    edit /etc/postfix/mail.cf to point to /etc/postfix/canonical with

      canonical_maps = hash:/etc/postfix/canonical

    in case of permission problems, reset as follows (as root):

      chown -R postfix /private/var/spool/postfix
      chown root /private/var/spool/postfix
      chown root /private/var/spool/postfix
      chown :postdrop /private/var/spool/postfix/public
      chown :postdrop /private/var/spool/postfix/maildrop
      chown :postdrop /usr/sbin/postqueue
      chown :postdrop /usr/sbin/postdrop
      postfix start      

  man procmail
  man procmailrc
  man procmailex

  $HOME/.forward should contain "|exec /usr/bin/procmail"  

ps aux                   on Sun machines
ps -efl                  on HP machines  

logical OR match in Emacs     \(foo\|bar\)  

• to extract a .spec file from a .rpm file:

   rpm2cpio foo.src.rpm | cpio -idm '*.spec'    

• RedHat: to turn off:

   /sbin/chkconfig sendmail off
   /etc/init.d/sendmail stop
   /etc/init.d/sendmail status    

• to install spamassassin itself

   * download tarball and unpack
   * perl Makefile.PL PREFIX=$HOME/local/i686/spamassassin-3.1.7
   * make
   * make install

   * edit $pfx/share/spamassassin/10_misc.cf replace this:

     add_header all Status "_YESNO_, score=_SCORE_ required=_REQD_ tests=_TESTS_ autolearn=_AUTOLEARN_ version=_VERSION_"

     with this:

     add_header all Status "_YESNO_, score=_SCORE_ required=_REQD_ tests=_TESTSSCORES_ autolearn=_AUTOLEARN_ version=_VERSION_"    

• to get the NETBL tests

   had to install Net::DNS and Net::IP locally in order to have a
   recent-enough version of Net::DNS

   - go to search.cpan.org, search for Net::DNS and Net::IP and download

   - unpack, then

     * perl Makefile.PL --noxs PREFIX=/path/to/spamassassin-3.0
     * make
     * make test
     * make install

     NOTE: when running 'make test' for Net-DNS, it will need to find
     the just-installed Net-IP module, so it needs to see that
     directory in @INC, which can be done by setting the PERL5LIB
     environment variable like this:

     export PERL5LIB=/home/rjpeters/local/i686/spamassassin-3.1.7/lib/perl5/site_perl/5.8.5

   - hand-edit /path/to/spamassassin-3.0/bin/spamassassin to have this
     line:

     use lib '/path/to/spamassassin-3.0/lib/perl5/site_perl/5.8.5';

     since by default it only referred to 5.8.3

   - edit ~/.procmailrc to point to the local config files:

   :0 fw
   | $SPAMASSASSIN -D --configpath=$SA_PFX/share/spamassassin --siteconfigpath=$SA_PFX/etc/mail/spamassassin 2>> $MAILDIR/sa.log


   - debugging by

     * adding -D to spamassassin cmd line, this showed that Net::DNS
       had too old of a version

   - reenable DCC in $PFX/etc/mail/spamassassin/v310.pre    

• To use sa-update

   just run sa-update (optionally with -D for debugging), but no need
   to give an --updatedir option; sa-update will automatically put the
   updates under $PFX/var/spamassassin... if $PFX is non-default

Sys::Hostname::Long
Net::CIDR::Lite
Mail::SPF::Query
IP::Country
IO::Zlib
Archive::Tar

http://wiki.apache.org/spamassassin/InstallingRazor    

• shorter line length with 'fold_headers' configuration option

   hand edited /path/to/spamassassin-3.0/lib/perl5/site_perl/5.8.3/Mail/SpamAssassin/PerMsgStatus.pm
   like this diff:

         - $Text::Wrap::columns = 79;
         + $Text::Wrap::columns = 70;    

• tunneling for VNC vnc

   setup: iMac is behind Linksys firewall/router
          firewall is set to forward port 22 to iMac
          iMac has its own firewall with port 22 open for ssh traffic

   to run VNC server on iMac and VNC viewer on some other internet client:

   client$  ssh -l username server_hostname -L 5900:127.0.0.1:5900
   server$  cd /Applications/OSXvnc.app/Contents/MacOS/
   server$  ./OSXvnc-server

       2004-08-04 13:55:24.713 OSXvnc-server[26405] Main Bundle: /Applications/OSXvnc.app
       2004-08-04 13:55:24.739 OSXvnc-server[26405] Waiting for clients
       2004-08-04 13:55:24.739 OSXvnc-server[26405] Started Listener Thread on port 5900

   so we ssh'd from the client into the server with port 5900 from the
   server being forwarded to port 5900 on the localhost
   (i.e. 127.0.0.1)

   now, from another terminal window on the client:

   client$  vncviewer 127.0.0.1:0

       VNC viewer for X version 4.0b4 - built Nov 17 2003 13:17:08
       Copyright (C) 2002-2003 RealVNC Ltd.
       See http://www.realvnc.com for information on VNC.

       Wed Aug  4 14:05:33 2004
        CConn:       connected to host 127.0.0.1 port 5900
        CConnection: Server supports RFB protocol version 3.3
        CConnection: Using RFB protocol version 3.3
        TXImage:     Using default colormap and visual, TrueColor, depth 24.
        CConn:       Using pixel format depth 6 (8bpp) rgb222
        CConn:       Using ZRLE encoding    

• ssh automatic authentication without password
• on client (locke)
• run ssh-keygen to create keys

    in ~/.ssh

    ssh-keygen -t rsa1 -f identity
    ssh-keygen -t rsa -f id_rsa
    ssh-keygen -t dsa -f id_dsa        

• on host (hume)
• setup host:~/.ssh/authorized_keys2

    copy literal contents of
      client:~/.ssh/identity.pub
      client:~/.ssh/id_rsa.pub
      client:~/.ssh/id_dsa.pub
    into host:~/.ssh/authorized_keys2        

• setup host:~/.ssh/config

    add to ~/.ssh/config
      PubkeyAuthentication yes
      RhostsRSAAuthentication yes
      HostbasedAuthentication yes        

• setup host:/etc/ssh/sshd_config

    add to /etc/ssh/sshd_config
      RhostsRSAAuthentication yes
      PubkeyAuthentication yes        

• on client (locke)

   make sure to "chmod og-rwx ~/.ssh"

   eval `ssh-agent`
   ssh-add ~/.ssh/id_rsa (enter passphrase)      

• scp between two remote machines
• Punchline: need to turn on agent forwarding, by putting the following

    line in ~/.ssh/config:

      ForwardAgent yes

    Here's an explanation from a usenet thread:

    http://groups.google.com/groups?hl=en&lr=&ie=UTF-8&threadm=Xns92AAD5C8420EAcelfers%40127.0.0.1&rnum=6&prev=/groups%3Fq%3Dscp%2Blocalhost%26meta%3Dsite%253Dgroups

    Say the host you are running the initial scp is A, and you're
    doing scp B:foo C:bar. Your command has the effect of running a
    second copy of ssh on B. It needs authentication information to
    login to C; how will it get it? Password authentication will not
    work at all, since scp on A invokes ssh with -n and provides no
    pty for the exec channel, so ssh on B can't prompt you for
    it. Public-key will be OK, but you have to use agent forwarding;
    again, even if an appropriate key is on B, it can't prompt you for
    a passphrase.      

• recursively copy a directory by piping tar through ssh

   This works better than scp, since scp does not respect and preserve
   symlinks, for example..

   localhost# tar cvfz - {somedir} | ssh {remotehost} "cd dest; tar xfz -"    

  single instruction multiple data

  MMX == SIMD integer operations
  SSE == SIMD floating-point operations

  MMX has 8 x 64-bit integer registers (borrowed from floating-point stack)
    mm0..mm7

  SSE has 8 x 128-bit floating-point registers
    xmm0..xmm7

  SSE2 was introduced with Pentium-4 and allows operations on 64-bit
  floating point operands  

  To boot into single-user mode in linux, select the desired kernel in
  GRUB but press "a" instead of Enter. Then, append the desired
  runlevel to the end of the boot parameters line, then press Enter to
  continue booting. E.g., for single-user mode, append a "1", and for
  text-only mode, append a "3".  

  To control from the command-line: use mdutil and mdimport.  

  http://www.brunolinux.com/01-First_Things_To_Know/Skinny_Elephants.html
  http://hehe2.net/linux-general/the-7-habits-of-highly-effective-linux-users/

  If all is blocked and even Ctrl+Alt+backspace (what should be a last
  resort) does not react, if your system does not react on any action
  remember the next line:

  Raising Skinny Elephants Is Utterly Boring

  Here is how you "raise the elephant":

  Alt+SysRq+r  ( The LEFT Alt key ) ( SysRq is on the same button as print screen )
  Alt+SysRq+s
  Alt+SysRq+e
  Alt+SysRq+i
  Alt+SysRq+u
  Alt+SysRq+b

  Give a little time between keystrokes.

  The r stands for put keyboard in raw mode
  The s for sync the disk
  The e for terminate all processes
  The i for kill all processes
  The u for remount all filesystems read only
  The b for reboot the system

  THIS IS THE VERY LAST SAVE YOUR BUTT PROCEDURE ! ONLY IF ALL ELSE FAILS !

  NOTE: For the skinny elephants to work you need to have the
  sysrq-key enabled in the kernel. (CONFIG_MAGIC_SYSRQ)  

• shell script workaround

#!/bin/sh
# \
exec tclsh "$0" "$@"    

• http authentication

    >  Now, what do you put in extraHeadersList? Something like:
    >set extraHeadersList "Authorization {Basic gqQls2l0daE60Gq9a39sBGE=}"

    The encoding scheme is simple, too.

    Basic authentication is just the user name and password, seperated by
    a colon, and base64 (MIME) encoded.

    For example, the encoded string for user "john" with password "today"
    would the Base64 encoded string of "john:today".    

  file delete $f --> unlink $f

  file tail $f --> use File::Basename; basename($f);

  file copy $f $g --> use File::Copy; copy($f, $g);

  file rename $f $g --> rename($f, $g);

  file extension $f --> use File::Basename; my ($name, $path, $ext) = fileparse($f, '\..*');

  file size $f --> (-s $f)

  file executable $f --> (-x $f)

  file exists $f --> (-f $f) or (-e $f)

  file isdirectory $f --> (-d $f)

  file mkdir $f --> mkdir($f)  

touch dt ; sudo date -s "`date -r dt`"; rm dt  

  date +%Y%m%d-%H%M%S  

• check current filesystem against tripwire database

   tripwire --check    

• update the database to match a tripwire report

   tripwire --update --twrfile /var/lib/tripwire/report/$HOSTNAME-$DATE-$TIME.twr    

• print a report

   twprint --print-report --twrfile /var/lib/tripwire/report/$HOSTNAME-$DATE-$TIME.twr    

lsusb; # to check if usb device was detected
sudo mount /dev/sda1 ~/cam  

• Commands to move around the file:

   h      Move the cursor left one character.
   j      Move the cursor down one line.
   k      Move the cursor up one line.
   l      Move the cursor right one character.
   <arrows> The cursor arrow keys should work, too.    

• Commands to enter new text:

   a      Append new text, after the cursor.
   i      Insert new text, before the cursor.
   o      Open a new line below the line the cursor is on; start entering text.
   O      Open a new line above the line the cursor is on; start entering text.
   <esc>  Quit entering text and return to command mode.    

• Commands to delete text:

   dd     Delete the line the cursor is on.
   x      Delete the character the cursor is on.    

• Commands to write the file:

   :w<return>             Write the file with its original name.
   :w file_name<return>   Write the file with the name "file_name".    

• Commands to quit the editor:

   :q<return>             Quit editing and leave vi.
   :q!<return>            Quit, discarding any modifications.    

how to just get the saliency map:

vision --just-save --save-salmap --exit-after-display foo.pgm  

Yech. Use ktnef or tnef to extract the contents.  

  dual-boot with Fedora

  Fedora changes the MBR so that the XP install CD barfs

  to get the XP install CD to work, need to boot into linux with a
  linux rescue CD, then DON'T mount the hard drive, then zero out the
  MBR like this:

    dd if=/dev/zero of=/dev/hda bs=512 count=1

  Then reboot with the XP install CD and install XP.

  Save the mbr with

    dd if=/dev/hda of=/boot/mbr.img bs=512 count=1

  It can then be restored with:

    dd if=/boot/mbr.img of=/dev/hda bs=512 count=1

  or if you do not want/need to overwrite the partition table with:

    dd if=/boot/mbr.img of=/dev/hda bs=446 count=1

  as the partition table is kept in the last 66 bytes of the MBR.  

• caps lock

   xmodmap -e "keycode 66 = "  # to disable caps lock
   xmodmap -e "keycode 66 = Caps_Lock"  # to re-enable caps lock
   xmodmap --help
   xev
   loadkeys    

• fonts

   -maker-family-weight-style-spacing--pixel size-tenth points-horz res-vert res-mono/prop-ave width-charset

   maker:        adobe, etc.
   family:       courier, helvetica, etc.
   weight:       medium, bold
   style:        (r)oman, (i)talic, (o)blique
   spacing:      normal, condensed, narrow, double

   pixel size:   font size in pixels
   tenth points: font size in tenths of a point
   horz res:     horizontal resolution for which font was designed
   vert res:     vertical resolution for which font was designed
   mono/prop:    (m)onospace, (p)roportional spacing
   ave width:    average width in tenths of a pixel
   charset:      character set of font, e.g. iso8859-1    

• key auto-repeat

   xset q            # to query current user preferences
   xset r on         # to turn on auto-repeat
   xset r off        # to turn off auto-repeat    

  find . -name \*.log | xargs -n 10 echo foo bar

  gives

  foo bar a.log b.log c.log d.log e.log f.log g.log h.log i.log j.log
  foo bar 1.log 2.log 3.log 4.log 5.log 6.log 7.log 8.log 9.log 10.log
  etc.  

• To make a data cd:
• select some tmp storage space in the top-level setup dialog
• go to "master tracks"
• in "master source", choose the directory tree to be mastered
• in "create session/image":
• first "calculate size"
• then "master to image file"
• go to "write tracks"
• set up a layout in "layout tracks", then "accept track layout"
• go to "write tracks" and choose the cd type, then hit "write tracks"