Repository navigation
Added
- CI validates one sample per PDF/A and PDF/UA claim with veraPDF, independently of the library's
own validators.
Changed
- Redaction now refuses with
PdfRedactionExceptionwhen content cannot be checked against the
regions (unresolved font, Form XObjects nested past 12 levels, or a form that draws itself)
instead of returning normally with that content left in place.
Fixed
-
Text extraction from Standard 14 Symbol fonts without explicit encoding now uses the built-in
Greek and mathematical character mapping instead of Latin StandardEncoding. -
Embedded TrueType CID fonts now declare empty glyph slots retained by subsetting in
CIDSet,
fixing independent PDF/A-2/3 and PDF/UA-1 validation failures. -
PDF/UA-2 declarations select PDF 2.0 and supply a missing namespace on a single
Documentroot;
the validator now rejects the wrong version and root namespace. -
Redaction removes
ActualText,AltandEfrom marked-content sequences and structure
elements whose content it touched; they used to keep the removed text extractable.
PdfRedactionResult.SemanticEntriesRemovedreports how many went. -
Redacting a Form XObject affects only the selected placement. Other pages and other placements
of the same form keep the original, and image entries are kept while another page draws them. -
DocMDP and FieldMDP verification compares each page's resource graph, so changing a Form
XObject, image or font behind an unchanged content stream is reported as a modification. -
An image whose
/SMaskor/Maskrefers back to itself raisesPdfFormatExceptioninstead of
overflowing the stack and ending the process; over-long mask chains raise
PdfResourceLimitException.
Validated with .NET SDK 10.0.403. Package API compatibility, Source Link, clean-cache consumers and Native AOT smoke tests passed. See attached SHA256SUMS for the exact package and symbol artifacts.