Prevent embedded file streams from being bytes

[Tontyna]

Should fix #644

[liZe]

As

• pdf_file_object.stream is initially an empty str
• decode always returns str objects
• adding str and bytes objects raises an exception,

I don't know how we could ever get a bytes object here!

[Tontyna]

I don't know whether we ever get a bytes object from decode (zlib is built-in, cannot debug), it was just an idea. It was just an experiment. And by forcing the stream to bytes the embedded stream in the PDF was changed into b'....', exactly what the buggy PDFs contain.

No exceptions here. Neither with pdf_file_object.stream = str(pdf_file_object.stream) nor with pdf_file_object.stream = bytes(pdf_file_object.stream, 'latin-1').

BTW: My pdfrw version is 0.4

[liZe]

I don't know whether we ever get a bytes object from decode (zlib is built-in, cannot debug), it was just an idea. It was just an experiment.

No problem 😉.

The bug is in 0.42.3 but is already fixed in master. I've backported the fix into the 0.x branch (as written in #644) and I'll release a new 0.x version later.

[Tontyna]

Ah, got it -- I (of course?) used the master branch when exploring the issue...