We no longer use Ghostscript to manage PDF metadata, instead
omitting the DOCINFO segment from the pdfmark file we generate.
Instead all of the relevant metadata code has been migrated to pikepdf,
and we use that API. This should be more consistent and fixes the
Ghostscript version-depedent quirks.
Also removes our python-xmp-toolkit dependency, except for
testing.
Ghostscript no longer supports UTF-16-BE-hex strings as a way of
supplying Unicode data in pdfmark so we have lost this functionality too:
http://git.ghostscript.com/?p=ghostpdl.git;a=commit;h=e997c6836d243ab37fe3a5f0d57974af95eb5eac
For users this means setting --title, --author, etc. will not work if gs
9.24 is installed, but if the file has existing metadata it might work.
For now we enforce police-state-strict ASCII, until there's time to
implement proper metadata editing. Relevant tests set to xfail.
Kodak Capture Desktop and probably other software creates a
/Outlines entry with /First being set to an invalid indirect reference to
an object that hasn't been created. This is legal in the PDF spec but
problematic for qpdf. The objgen will be (max valid object ID + 1, 0).
Because we create new objects in _weave, some TOC entries will end
up assigned to new objects we create. Typically /ProcSet.
We solve the issue by refactoring page traversal and then doing it
twice, once to resolve all references (eliminating the null
reference problem) and a second pass to make our changes.
For some reason PyPDF2 has begun to trigger internal errors in
pytest on macOS alone. Not sure why, but nothing is wrong that I can
see. Seemed like an opportune time to switch to pikepdf; found some
new issues in the process anyway.
Related to previous validation issue. If the /CreationDate had no
timezone, Ghostscript also creates invalid metadata. Work around this.
Also fix up PDF date decoding, and transcode dates to standardize them.
