FDF is not properly cleaned up on pdftk errors

[suspectpart]

Working with your library extensively today (it's great!), I noticed that whenever the pdftk subprocess crashes, some artifacts are not being cleaned up from /tmp.

Here is a proof that asserts that in case of a pdftk crash, the generated FDF remains in /tmp:

import os
import re
import subprocess

from pypdftk import fill_form

datas = {
    # this is gonna crash pdftk, because the angle bracket is not escaped
    'Given Name Text Box': '1 < 2'
}

try:
    fill_form('./OoPdfFormExample.pdf', datas=datas, out_file='out.pdf')
except subprocess.CalledProcessError as e:
    fdf_file = re.match(r'^.+(/tmp/.+?)\s', str(e)).group(1)

    assert os.path.exists(fdf_file)

(The PDF sample is taken from this source)

An opening angle bracket makes pdftk crash, because it's rendered into the XML as-is and not escaped (that's an issue of it's own - gonna file that too).

The fill_form function looks like this:

def fill_form(pdf_path, datas={}, out_file=None, flatten=True):
    cleanOnFail = False
    tmp_fdf = gen_xfdf(datas)
    handle = None
    if not out_file:
        cleanOnFail = True
        handle, out_file = tempfile.mkstemp()
    ...
    try:
        run_command(cmd, True)
    except:
        if cleanOnFail:
            os.remove(tmp_fdf)
        raise
    finally:
        if handle:
            os.close(handle)
    os.remove(tmp_fdf)
    return out_file

If run_command(cmd, True) raises, first the except, then the finally block are executed. Because the except block re-raises, the last two lines (after the finally) block are never executed. So in case cleanOnFail is False (when outfile is being passed as an argument), the tmp_fdf is never removed.

There is another bug hiding in this: If outfile is not being passed, a temporary file is created and a handle is opened:

 if not out_file:
        cleanOnFail = True
        handle, out_file = tempfile.mkstemp()

In the finally block, the handle is closed, but the out_file is not being removed (although the failing pdftk already created the file, even though it's 0 bytes. It almost looks like a copy-paste problem, that if cleanOnFail is True, actually os.remove(out_file) should have been called?

I can fix both problems and create a PR, if you like.