I'm converting a HTML file to PDF using the following command:

~/Downloads/prince/bin/prince /Users/fnando/Sites/howto-books/ruby-metaprogramming/output/ruby-metaprogramming.html -o /Users/fnando/Sites/howto-books/ruby-metaprogramming/output/ruby-metaprogramming.pdf

Sometimes it works, sometimes it does not. Here's the output from several tries. As you can see, there's no defined pattern.

http://pastie.textmate.org/private/9hye0cq2jcn61gthutvkuw

Do you have any workaround? I tried with both Prince 7.1 and 8.0 beta 1 on Mac OS X 10.6.7 without success.

Ouch, that's nasty. To figure this out it would help to make the input document as small as possible while still replicating the error. Does the document reference any images? Does it download linked resources over HTTP?

It does reference local images (both PNG and SVG). I'll try to get something that reproduces this error asap.

Commenting out images is a good start, disabling style sheets one by one, using fewer fonts, chopping the document in half, and so on. It is complicated slightly if the error doesn't happen every time, though.