Files
ioccc/2012/hamano
Cody Boone Ferguson 3deb7ae603 Improve markdown guidelines wording
Amongst the changes here is that we do not need to have the hex or html
code (numbers) for open and close parentheses (in links) as there are
names. '(' is & followed by lpar; and the ')' is the same but 'rpar;'
instead. This makes it easier to remember and type and it generates the
same html as well (they are all valid). Now the markdown files are
consistent in this (in some cases hex was used and in others the html
code).

Although many markdown files were updated no html files needed to be
updated as it generates the same html. Even so I ran the appropriate
rules to do so just to verify this.
2024-07-31 10:44:10 -07:00
..
2024-06-27 06:15:35 -07:00
2020-12-30 16:57:03 -08:00
2023-07-06 21:49:38 -07:00
2024-07-31 10:44:10 -07:00

To build:

    make

To use:

    ./hamano < textfile > output.pdf

Try:

    ./try.sh

The script will generate a number of PDF files and it will suggest you look at them as well. You can read hint.pdf with a PDF reader like acroread hint.pdf or evince hint.pdf. With macOS you can do open hint.pdf which by default will open Preview.app. You may of course use your pdf viewer to open the file directly instead.

One of the PDF files generated the README.md file as a PDF, obfuscated, and the PDF file will then be compiled as C (itself!) and executed so that it shows README.md!

That procedure looks like:

    ./hamano < README.md > hint.pdf
    cc -xc hint.pdf -o hint
    ./hint

Another PDF generated is an obfuscated Hello World! program (obfuscated inside the PDF). Once this is done it will compile the PDF as if it was C (itself!). It will look like:

    echo 'int main(){puts("Hello World!");}' | ./hamano > hello.pdf
    cc -xc hello.pdf -o hello
    ./hello | cc -Wno-implicit-function-declaration -xc - -o ./hello2

    ./hello2

although the CFLAGS will be what is in the Makefile as it uses the helper rules hint.pdf, hint, hello.pdf and hello to do this.

Judges' remarks:

This entry treads into a new territory for IOCCC - generating PDF files.

The originally submitted entry could have been thought of as a "Best abuse of Ghostscript" entry. :-)

The source code and rendered pages are obfuscated for humans (where is the font?), and the submitted entry created PDF files that left Ghostscript confused.

The results can be viewed with Acrobat Reader, evince and Apple's Preview.

The updated version can also be viewed using Ghostscript.

There is something strange about the little flags in the output. A search of English literature will provide a hint. The conclusion you might draw is too elementary. :-)

So what do those flags really mean?

The Dancing men algorithm might be useful.

Don't forget to take a look at the generated PDF. Perhaps you might even want to compile the output with a C compiler. When you run it, what does it output?

Author's remarks:

This program obfuscate text file into PDF file with Dancing men algorithm.

Probably, the output PDF file is compliant with PDF 1.3 and it's also available to compile as C code.

I've tested it with GCC 4.7 and Clang 3.0 with Linux, and the following PDF Readers:

  • Adobe Acrobat Reader
  • Evince
  • Ghostscript
  • Xpdf

Obfuscations

This program is obfuscated by using classical methods. But can you find the embedded font in the fragmented glyph?