LibRaw 202502 snapshot is just published in our Github repository. Changes are:
Camera format support

• OM System 14-bit high-resolution files.
  Note: the new decoder supports old (12-bit) Olympus/OM-System files
  too and slightly faster (about 5%), so the old one is removed.
• Nikon NEFX (4/8/16/32 shots in PixelShift mode merged via Nikon software)

Camera support

• Canon EOS R5 C
• Hasselblad CFV-50c, CFV-100c
• Leica SL3
• Nikon Z f (standard/lossless compression only)
• Olympus/OM System OM-1 Mark II, TG-7
• Pentax KF, K III Monochrome
• Samsung S23+, S23 Ultra
• Sony A9-III, ILX-LR1

API Changes/improvements

• Exif callback is called from CR3 parser before calling parse_makernotes()
  parameters are the same as usual, except tag parameter value.
  tag values:
  (0x70000 | 0x927c) - if called from CMT3 tag
  (__type << 20) | 0x80000 | 0x927c - if called from CTMD record with type __type
• Exif callback is called for Sony SR2 tag with tag parameter set to 0x60000 | tagID
• Makernotes callback implemented for TIFF-like (IFD structure makernotes)
  Callback is very similar to exif_callback.
  To set callback use LibRaw::set_makernotes_handler(exif_parser_callback cb, void *data)
  Callback is called (if set) with tag value set to (uptag << 16)|makernotes_tag
  where uptag is
  • EXIF tag: if called from EXIF makernotes parser
  • 0: if called from CR3 metadata parser
  • tag-ID if parse_makernotes() is called from some specific tag context (see parse_makernotes calls in source code).
• New metadata field makernotes.pentax.DynamicRangeExpansion[4]
  if DynamicRangeExpansion[1] is non-zero, DynamicRangeExpansion[0] should be added
  to BaselineExposure
• New API call (static) int LibRaw::simplify_make_model(unsigned * mkindex, char *mkbuf, unsigned mkbuf_size, char *modelbuf, unsigned modelbuf_size):
  This call can be used to simplify/standardize vendor name/model received
  from external source (e.g. EXIF block in JPEG file).
  mkindex: pointer to received maker index (LibRaw_CAMERAMAKER_...)
  mkbuf, mkbuf_size: buffer with make string, buffer size
  modelbuf, modelbuf_size: buffer with model string.
  simplify_make_model(..) simplifies/re-writes vendor name/model in place, in provided buffers.
  return values:
  0 - mkindex is set (if non-null pointer provided), data simplified;
  1 - mkindex is not set, while some data processing is done;
  -1 - incorrect parameters;
• All file offsets are 64-bit now;
  Note: data callback signature has changed
• New LibRaw_abstract_datastream (and derived classes) methods:
  bool is_buffered() - returns current buffering mode (if implemented);
  void buffering_on() - turns buffering on.
  These methods complement the (existing) buffering_off() method
• New bits in imgdata.process_warnings:
  LIBRAW_WARN_DNG_NOT_PROCESSED - RAW file was passed to DNG SDK, but not processed.
  LIBRAW_WARN_DNG_NOT_PARSED - DNG parse error or non-valid DNG at valid_for_dngsdk() stage
• New compile-time define: LIBRAW_MAX_CR3_RAW_FILE_SIZE
  Default: value of LIBRAW_MAX_NONDNG_RAW_FILE_SIZE
  CR3 file size is limited by 4GB (max MP4 container size), so it may be needed
  to set the limit above non-DNG 2GB file size limit.

Minor improvements

• Adobe DNG SDK 1.7.1 compatibility
• restored compatibility with pre-1.7 Adobe DNG SDK
• RawSpeed1/3: do not pass files larger then 2GB
• Jpeg XL preview file writer
• Jpeg XL preview support in simple_dcraw -E/-e
• DNG 1.7 JpegXL DNG/GoPro DNG: do not refuse files at open_file() if LibRaw is not built with DNG SDK/GoPro SDK
  Note: LibRaw::get_decoder_info() will return decoder flag bit set to LIBRAW_DECODER_UNSUPPORTED_FORMAT if
  LibRaw is not built with corresponding SDK.
• All implicit type conversion *excluding INT64<=>int/uint* conversions are made explicit
  to remove C++ compiler type conversion warnings.
• libraw_dng_stream: underlying LibRaw_abstract_datastream buffering restored in
  destructor call.

Bug/problem fixes

• Fix for PanasonicV8 decoder compile with OpenMP
• Fix for incorrect 32-bit shift in Pana8 decoder generated by gcc11+
• DNG SDK Integration: clear imdata.color.linear_max[] if Stage3 opcodes are applied.
• Fixed possible buffer overrun in LibRaw::remove_trailing_spaces call
• CR3 decoder: workaround for possible incorrect generation of 64-bit code
• Improved Sony metadata processing
• fixed possible static table read overrun (if specially crafted image
  data passed to LibRaw)
• Fix for monochrome DNG files compressed as 2-color component LJPEG
• Fixed integer overflow in largest DNG frame selection code
• Additional out-of range checks to better handle specially crafted files
• Prevent out-of-buffer access in phase_one_correct()
• Prevent one-off read in Fuji WB table parsing
• Prevent one-off read in phase_one_correct()
• Do not apply Canon metadata crop values to DNG files
• Adjust profile_length to allocated data size.