PDF Tools SDK Release Notes

Added

• ForbidImplicitUpgrade option on ConversionOptions to prevent automatic PDF/A conformance level upgrades during conversion.

Changed

• HEIF/HEIC codec library dynamically linked

  The native library requires the HEIF/HEIC decoding library at load time, even if your application doesn’t use HEIF/HEIC functionality. If the decoding library isn’t found, the native library fails to load. The required binaries are included in the distribution, but ensure your service, container layout, library search paths, and permissions allow the library to be found.

  Action:
  After updating, verify that your deployment loads correctly even if HEIF/HEIC processing is not used.

• JBIG2 decoding update

  The JBIG2 decoding implementation has been replaced. JBIG2-encoded images render identically, but error messages related to JBIG2 processing may differ from previous versions.

  Action:
  If you process PDFs with JBIG2-encoded images, consider validating against a representative document set.

• Updated font substitution for missing fonts (including PDF/A)

  When PDFs reference fonts that aren’t embedded and are unavailable on the host, the software substitutes fallback fonts. The updated fallback fonts support a larger glyph set and better fitting to original glyph dimensions. Layout should remain stable, but visual appearance differs from both the original font and previous versions.

  Action:
  If exact reproduction is required, install the original font in your environment.

• ZapfDingbats font delivery change

  ZapfDingbats is no longer shipped. The fallback font URW++ D050000L.ttf is used instead, but it doesn’t provide full coverage and may differ slightly in appearance.

  Action:
  If exact ZapfDingbats reproduction is required, install the font in your environment.

• Refreshed internal mapping tables

  Character mapping tables used for Unicode and glyph mapping have been updated. Changes are primarily additive (new code points) with occasional correctness fixes.

• Java: native library loaded automatically

  Before this release, Java applications had to explicitly locate and load the Pdftools SDK native binary using System.load() or System.loadLibrary() before calling any SDK function. With this release, the native binary is loaded automatically at runtime.

  The Maven dependency has also been simplified: the platform-specific native artifact classifier is no longer needed.

  Action:
  Remove any System.load() or System.loadLibrary() calls for the Pdftools SDK native library from your code. In your code pom.xml , replace the two-dependency block (main JAR + native artifact) with a single dependency on pdftools-sdk . Refer to the Java getting started guide for the updated configuration.

• Upgraded libxml2 from version 2.10.3 to 2.15.0. Error messages related to XML processing may differ from previous versions.

Fixed

• Debug-level log messages are written to the file output sink when file logging is configured via PDF_TRACE_OUTPUT or PDFT_LOG_OUTPUT.
• Fixed a crash when repairing PDFs with XMP metadata dates missing timezone information.
• PDF/A conversion no longer fails when the document contains a /Tabs entry written as a string instead of a name object.
• PDF/A conversion no longer fails when the document contains a CIDFont dictionary without a /Type entry.
• The remove-PieceInfo option (eStripPieceInfo ) also removes PieceInfo dictionaries from XObject resources.
• Before this update, on Linux, PDF dates were computed using the UTC offset of the current time instead of the UTC offset applicable to the date being processed. As a consequence, PDF dates could be off by one hour when the document’s date and the current time fell in different daylight saving time (DST) periods. With this release, the UTC offset is correctly derived from the date being processed.
• Before this update, the JBIG2 decoder placed no limit on memory allocations, which could cause excessive memory consumption when processing PDFs with malformed or unusually large JBIG2-encoded images. With this release, JBIG2 memory allocation is capped at 2 GB.
• Before this update, calling Extractor.ExtractText() using the C# binding on Linux produced incorrect output. As a consequence, non-ASCII UTF-8 characters were silently dropped from the extracted text. With this release, text extraction produces correct UTF-8 output on Linux.

Added

• Enhanced logging configuration environment variables:
  • PDFT_LOG_OUTPUT for output control, replacing PDF_TRACE_OUTPUT. The PDF_TRACE_OUTPUT also remains supported for backward compatibility.
  • PDFT_LOG for log level filters. The comma-separated format: *:error,libbase:info , where * sets the default level.

Changed

• Switched from the zlib to zlib-ng library, significantly improving performance when compressing and decompressing Flate streams by using SIMD instructions.

Fixed

• Corrected conformance retrieval logic in the GetConformance method to properly handle invalid ConversionOptions values.
• Certificate loading on Windows now includes Local Machine store in addition to Current User store, resolving certificate access issues for Internet Information Services (IIS) applications running under service accounts.
• Widget annotation appearance generation now correctly checks for rich text value (RV) entry instead of rich text style (DS) entry when determining whether to create appearance streams.

Fixed

• Resolved an issue using an uninitialized variable that could cause unpredictable behavior during PNG processing.
• Fixed a stack buffer overflow when creating annotation appearances.
• Corrected glyph position problems caused by improper handling of adjustment values.
• Ensured the GTS_PDFX output intent is correctly copied into the output document to maintain PDF/X compliance. The GTS_PDFX defines the standardized printing output intent for PDF/X files.
• Fixed an infinite loop that could occur during optimization for specific malformed PDF files.
• Addressed an access violation related to cross-reference (xref) tables, preventing potential crashes.

Added

• Added IsFullRevisionCovered flag to signature validation results to detect partial ByteRanges.

Changed

• Removed library load/unload messages from Python bindings.

Fixed

• Font name decoding for Type0 fonts on Linux and macOS to preserve Unicode characters.
• Malformed import in native_base.py that caused errors in Python bindings.
• Fixed error reporting for non-UTF8 colorant name on Linux and macOS.
• Font name decoding on Linux and macOS to align behavior with Windows.
• Fixed unexpected termination when processing image dictionaries with an empty /Filter array.
• Removed stray null terminators from strings returned by the Python interface.

Added

• With this release, we are launching a new Python Jupyter notebooks repository demonstrating the core functionality of the Pdftools SDK, including step-by-step examples for key features and use cases.

Fixed

• Previously, rendering PDFs with the wrong alternate colorspaces could cause the Pdftools SDK to crash. With this update, the rendering engine now falls back to CMYK rendering if the alternate color transform cannot be created, preventing the crash.
• Prior to this update, a crash occurred when IIS processes terminated, causing the Pdftools SDK DLL to unload without properly releasing its resources. With this release, the issue has been resolved, allowing the DLL to safely unload even when shutdown hooks fail to execute as expected.
• Previously, document timestamps (ETSI.RFC3161) were not recognized if the PDF lacked an explicit declaration of the correct signature type. This update ensures that timestamps are correctly validated as long as the signature itself is valid and the signed content remains intact.