How can a bash script write out a binary file, without using uudecode?

I've got a little puzzler here for the bash scripting experts... I have a bash script that needs to create a small (80 byte) binary file when it runs. The contents of the file need to be contained inside the script itself (i.e. I don't want to just package the file along with the script).

My script currently does it like this:

echo 'begin-base64 644 dummy.wav' > /tmp/dummy.uu
echo 'UklGRkgAAABXQVZFZm10IBAAAAADAAEAAHcBAADcBQAEACAAZmFjdAQAAAAAAAAAUEVBSxAAAAAB' >> /tmp/dummy.uu
echo 'AAAAQDYlTAAAAAAAAAAAZGF0YQAAAAA=' >> /tmp/dummy.uu
echo '====' >> /tmp/dummy.uu
uudecode -o /tmp/dummy.wav /tmp/dummy.uu
rm /tmp/dummy.uu

... after the above runs, I have my file /tmp/dummy.wav. But I just found开发者_开发百科 out that the computer this script is to run on does not have uudecode installed (and I'm not allowed to install it), so I need to find some other way to create this file. Any ideas?

Some installers do something similar to this:

#!/bin/bash
tail -n +4 $0 | tar xvzf -
exit
<tgz file appended here><newline>

If the target computer has perl available:

perl -ne 'print unpack("u",$_)' > dummy.wav <<EOD
M4DE&1D@```!7059%9FUT(!`````#``$``'<!``#<!0`$`"``9F%C=`0`````
C````4$5!2Q`````!````0#8E3```````````9&%T80``````
EOD

That's using the non-base64 format you get from just doing uuencode dummy.wav < dummy.wav on your original computer.

Failing that, you can always do this:

echo -ne '\x52\x49\x46\x46\x48\x00\x00\x00'  > dummy.wav
echo -ne '\x57\x41\x56\x45\x66\x6d\x74\x20' >> dummy.wav
echo -ne '\x10\x00\x00\x00\x03\x00\x01\x00' >> dummy.wav
echo -ne '\x00\x77\x01\x00\x00\xdc\x05\x00' >> dummy.wav
echo -ne '\x04\x00\x20\x00\x66\x61\x63\x74' >> dummy.wav
echo -ne '\x04\x00\x00\x00\x00\x00\x00\x00' >> dummy.wav
echo -ne '\x50\x45\x41\x4b\x10\x00\x00\x00' >> dummy.wav
echo -ne '\x01\x00\x00\x00\x40\x36\x25\x4c' >> dummy.wav
echo -ne '\x00\x00\x00\x00\x00\x00\x00\x00' >> dummy.wav
echo -ne '\x64\x61\x74\x61\x00\x00\x00\x00' >> dummy.wav

This little bit of bash was generated with:

$ hexdump -e '"echo -ne '\''" 8/1 "x%02x" "'\'' >> dummy.wav\n"' dummy.wav | sed 's;x;\\x;g;1s/>/ /'

Edited to add: As pointed out in a reply here, something like this is also a possibility:

xargs -d'\n' -n1 echo -ne > dummy.wav <<EOD
\x52\x49\x46\x46\x48\x00\x00\x00\x57\x41\x56\x45\x66\x6d\x74\x20
\x10\x00\x00\x00\x03\x00\x01\x00\x00\x77\x01\x00\x00\xdc\x05\x00
\x04\x00\x20\x00\x66\x61\x63\x74\x04\x00\x00\x00\x00\x00\x00\x00
\x50\x45\x41\x4b\x10\x00\x00\x00\x01\x00\x00\x00\x40\x36\x25\x4c
\x00\x00\x00\x00\x00\x00\x00\x00\x64\x61\x74\x61\x00\x00\x00\x00
EOD

(the -d argument is important to turn off xargs's own backslash processing)

You can also turn the 8/1 in my hexdump command into 80/1 and have a single long echo line.

Put an exit at the end of your script, append the file to the end of the script, and use tail -c 80 to get at the contents. This will work as long as you don't need to worry about newline conversion issues.

This is another example to decode radix 64 formatted data, it runs slow, but it is functional.

#!/bin/bash
exec<$0
while read line ; do if [ "$line" = "#payload" ] ; then break; fi; done
r64='ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/'
i=0; while [ $i -lt 256 ] ; do tab[$i]=-1 ; let i=$i+1 ;done
i=0; while [ $i -lt 64 ] ; do tab[`printf "%d" "'${r64:$i:1}"`]=$i ; let i=$i+1; done
bi=0
while read -n 1 x
do
 in=${tab[`printf "%d" "'$x"`]}
 if [ $in -ge 0 ]; then case $bi in
  0 ) out=$(($in<<2)); bi=6 ;;
  2 ) out=$(($out|$in)); printf \\$(printf '%03o' $(($out&255)) ); bi=0 ;;
  4 ) out=$(($out+($in>>2))); printf \\$(printf '%03o' $(($out&255)) );
  bi=0; out=$(($in<<6)); bi=2 ;;
  * ) out=$(($out+($in>>4))); printf \\$(printf '%03o' $(($out&255)) );
  bi=0; out=$(($in<<4)); bi=4 ;;
  esac fi
done
exit
#payload
dnZ2dnZ2dnZ2dnZ2dnZ2dnZ2XmAgICAte3Z2dnZ2dnZ2dnZ2dnZ2dgp2dnZ2dnZ2dnZ2dnZ2dnZ2PiAg
ICAgIC4gLXZ2dnZ2dnZ2dnZ2dnZ2CnZ2dnZ2dnZ2dnZ2dnZ2dn0gICAgICAgPT4gLXZ2dnZ2dnZ2dnZ2
dnYKdnZ2dnZJdnZJdnZJdnZJOyAgICAgICAtICAgPXZJdkl2dkl2dkl2dgp2dnZ2SXZ2dnZ2dnZ2dnZg
ICAgICAgICAgICAgbnZ2dnZJdnZ2dnZ2CnZ2dnZ2dnZ2SXZJdnZJdiAgIC4gICAgICwgICA8dnZ2SXZ2
dkl2dkkKdnZ2SXZ2SXZ2dnZ2SXZJIF9zOyAgX3VvLyAgID12dnZ2dnZ2dnZJdgp2dnZ2dkl2dnZJdnZ2
dnYgdyRtICBtQCRtICAgPXZ2dnZJdnZJdnZ2CnZ2dnZJdnZ2dnZ2dkl2SSBmPTQuO1cgYFE7ICA9dnZ2
dnZ2dnZ2dnYKdnZ2SXZ2dnZJdnZJdnZ2IHQtM3MlJiAgbWAgID12dnZ2SXZJdnZJdgp2dnZ2dnZ2SXZ2
dnZ2dnYgXWlvWjZYYXVQICAgPXZ2dnZ2dnZ2SXZ2CnZ2dkl2dkl2dnZJdnZJdi4pbVojWlojWlMgICAu
dnZ2SXZJdnZ2dnYKdnZ2dnZ2dnZ2dnZ2SXZ2OjNYWlpaI1pTWCAgICB7dnZ2dnZ2dkl2dgp2dnZ2SXZ2
SXZ2SXZ2dnY7PFNYWlhTZFhuIC5pLj12dnZJdnZJdnZ2CnZ2dkl2dnZ2dkl2dnZ2dmBdJVhYWlhubW0+
IC1gIHZ2dnZ2dnZ2dnYKdnZ2dnZ2SXZ2dnZ2SXYlIGptdklud1FXUW0gICAgPHZ2SXZ2SXZ2SQp2dnZJ
dnZ2dkl2dkl2dmAuUVFvWG1tUVFRUWMgICAge0l2dnZ2dnZ2CnZ2dnZ2dkl2dnZ2dnYrIGpRV1FtV1FR
UVFRayAgICAtdnZ2dkl2dkkKdnZ2dkl2dnZ2SXZJPiBfUVFRUVFRUVFRUVFRLiAgICA9dkl2dnZ2dgp2
dnZJdnZ2SXZ2dmwgIF1RUVFRV1FRUVdXOCRMICAgICA8dnZ2SXZ2CnZ2dnZ2dnZ2dnZ2OyAgbm1RUVFt
UVFRbXdvb20gLiAgIC1JdnZ2dnYKdnZ2SXZ2SXZ2SX0gID1RV1FRUVFRUVFRUVFtMlsgLSAgID12dkl2
dgp2dnZ2dnZ2dkl2Oy4gZFFRUVFRUVFRUVFRUVFRcSAgLiAgIEl2dnZ2CnZ2dnZJdkl2dnZgLjxRUVFR
UVFRUVFRUVFRUVdRKC4uLiAgPEl2dnYKdnZ2SXZ2dnZ2PiAgZFFRUVFRUVFRUVFRUVFRUVFbICAuICAg
dnZ2SQp2dnZ2dnZ2dnYnIC5RUVFRUVFRUVFRUVFRUVFRUWsgIC4gICB7dnZ2CnZ2dkl2dkl2PiAuXVFR
UVFRV1dXUVFRUVFRUVFRbSAgICAgIClsdnYKdnZ2dnZ2dnZgIDpqUVFRUVEjUVdRUVFRUVFRUVFXICAu
ICAgOnZ2SQp2dnZ2SXZ2bCAgOmpXUVFRUUVXV1FRUVFRV1FRUVcgIGAgICA6dnZ2CnZ2dkl2dnZJLl86
alFRUVFRRVdRUVFRUVFRUVFRVyAuIC4uID12dnYKdnZ2dnZ2dnZkIzYvUVdRUVFFUVFRUVFRUVFRV1dM
LiAgIDogKXZ2dgp2dnZJdnZJMyNaWkwtJFFRUVFRV1FRUVFRUVFCWiNgICAgLmRvdnZ2CnZ2dnZ2SXZa
IyMjWj4tNFFRUVdRUVFRUVFRUUVaay4gICBqWlh2dnYKdnZ2dndvbVgjWiNaIy4gNFFRUVFRUVFRUVdX
MVpYc189dVhaaHZ2dgp2dnZaWiNaI1VVWiNaTCAgXVFRUVFRUVFRUVdlWFpYcVhtWiNVenZ2CnZ2SVgj
I1ojWiMjWiNaLyAuUVFRUVFRUVFRVzEzI1paWlojWiMjb3YKdnZ2ZFVaIyNVWiMjVVVoX2FRUVFRUVFR
UVFQOlhaIyNVI1ojVVojaAp2dklkIyNaI1ojI1ojWlpaV1FRUVFRUVFXUCA9ZFojWiNaIyNaIyNaCnZ2
dlojWiMjVVVaI1ojWlpKUVFRUVFXUF4gIClYIyNaI1VVWiNVWjEKdnZ7WlojWlVaIyNaIyNaVXMtIT8i
fiAgICAgdlgjWiMjWiNaWF5sdgp2bCBZWFhYWFpaVVUjWlpaMS4gICAgICAgICB2WFojWiNaWCIgIDx2
CnZzICAtfiJJMVhYWFpaWm8xICAgICAgICAgIEluWFhaU31gICAgPHYKdnY7ICAgICAtLTwxMjIxbGAg
ICAgICAgICAgPElubjF9ICAgICB2SQp2dmwsICAgICAgICB+Kz5gICAgICAgICAgICAgfnwrfiAgICAu
JUl2CnZ2dnZpLiAgICAgICAgICAgIF9pc2ksICAgICAgICAgICAgX3ZJdnYKdnZ2dnZ2c19fXy4uLi5f
XyV2dnZ2SXZpLCxfLiAuLl9fPnZ2dnZ2dgp2dnZ2SXZ2dm52dnZ2dnZudnZ2dnZ2dnZubnZ2dnZ2dnZ2
dnZ2dnZ2Cg==

#!/bin/bash

# Define usage help
usage () {
    echo -e "USAGE:\n\t$0 <file to create> <dir to tar> <name of script or command to run>\n"
    exit 0
}

# check commandline arguments
if [ "$1" = "-h" ]; then usage; fi
if [ -z $1 ]; then usage; fi
if [ -z $2 ]; then usage; fi
if [ -z $3 ]; then usage; fi

# test for the directory and if it exists, create the bin file using tar
if [ -d "$2" ]; then
    cat >$1<<EOF
#!/bin/sh -e
sed -e '1,/^exit$/d' "\$0" | tar xzf - && "./$2/$3"
exit
EOF
    tar czf - $2 >> $1
else
    echo "$2 does not exist, aborting!"
    exit 1
fi

# make the new file executable and exit
chmod +x $1
exit 0

I would use base64 encoding, as that seems to be the general replacement for uu encoding, and operates on very similar principles.

From my point of view uuencode and uudecode are essential, but that's only my opinion. Without creating temporary files you could also do something like this (uudecode.sh):

#!/bin/bash
# --
# -- Uudecoding without using a regular temporary file
# --

# -- Create a named pipe:
mknod /tmp/.dummypipe p

# -- Starting uudecoding on that pipe in the background:
uudecode -o dummy.txt /tmp/.dummypipe &

# -- Push base64-uuencoded content into the named pipe:
cat <<END_DUMMY > /tmp/.dummypipe
begin-base64 644 dummy.txt
VGhpcyBpcyB0aGUgdXVkZWNvZGVkIHRleHQuCg==
====
END_DUMMY

# -- Remove the named pipe
rm /tmp/.dummypipe

Just encode the binary data in base64 and do something like this:

#! /bin/bash

cat <<EOF | base64 -d > wherever-it-goes.bin
UtEOtUaZcUCrEJtPJrA34BH8Wdpxb1/DtfMo5syiE/h+moNyApEkg2ZwA2/jTDdfl4WijnNbMzvZ
RrF3i7X353AjmTjLBz1NcOOJJhRPYLJ4WQYONyYj/fAhnXQd+s4SHaNponOWKj1AAzdlJY1VLWaX
P8QBJQcn2FTL4pJ3N04=
EOF