Follow-on to the non-dictionary /Font and /XObject guard. Replace the
_dict_entries() helper with an isinstance(..., pikepdf.Dictionary) check
at each resource lookup, which pikepdf's metaclass supports directly and
which reads as exactly the invariant being enforced. Iterate via
as_dict().values() so the values are typed and mypy stays clean once the
Any from the untyped resources argument is narrowed away.
Also guard _cid_font_is_embedded against a non-dictionary /FontDescriptor:
`key in descriptor` raises ValueError on a non-dict, which the caller's
except (AttributeError, TypeError, KeyError) does not catch. Such a font
now counts as non-embedded and is reported, so PDF/A conversion is refused
rather than risking Ghostscript corrupting a pre-existing CID text layer.
Adds a regression test for the FontDescriptor case (issue #1713).
A malformed PDF can store a non-dictionary object under a page's /Font or
/XObject resource. find_nonembedded_cid_fonts() iterated .values() on that
object outside the per-entry try/except, so scanning such a page raised
(TypeError/AttributeError depending on the pikepdf version) instead of
producing output. This surfaced as a PDF/A conversion crash.
Route both resource lookups through a small helper that returns an empty
list when the resource is missing or not a dictionary, so a garbage entry
is simply treated as having no fonts. Add a regression test covering a
non-dictionary /Font and /XObject.
Ghostscript's PDF/A conversion re-embeds non-embedded CID (CJK) fonts by
substituting a system font, which corrupts the character-to-Unicode
mapping and silently destroys an existing text layer -- commonly the OCR
layer Adobe Acrobat adds to scanned CJK documents.
Detect non-embedded CID/Type0 fonts before conversion: with
--output-type auto (the default) downgrade to a regular PDF and preserve
the text layer; with an explicit --output-type pdfa* stop with an error
rather than emit corrupted output. Simple non-embedded fonts (e.g. Latin)
are left alone -- Ghostscript substitutes them without corrupting the
text, and they are far too common to treat as conversion blockers.
Use --output-type pdf to keep the existing text layer, or --force-ocr to
rebuild it with embedded fonts.
Acrobat insists that PDF/A-1b should not have object streams.
Other programs like veraPDF disagree with this restriction, but
we can accommodate Acrobat so we will.
Also add more tests around this.
James R. Barlow
Matt Van Horn