Dear Maintainer,

Running pdftotext on the attached PDF produces some spurious form feeds.
I'm using pdftotext to extract text for indexing an 800-page document,
and the spurious form feeds throw off the page counts.

I've worked around the problem by altering libpoppler46 to grab the
page-termination sequence from an environment variable, but that's
just a hack.

  - A good fix would be to find out why these vertical bars are
    rendering as form feeds and to render them as suitable ASCII or
    Unicode characters.

  - An acceptable workaround would be to replace the -nopgbrk option
    with an option that sets the "page break" string to a string given
    on the command line (which could be empty).  That second solution
    would appear to involve some nontrivial refactoring of the code
    base, which is why I haven't tried to make a patch for it.