T A B L E o f C O N T E N T S --------------------------------- 1 Building Adobe XMPsdk and Samples in Terminal with the ./Generate_XXX_mac.sh scripts 1.1 Amazing Discovery 1 DumpFile is linked to libstdc++.6.dylib 1.2 Amazing Discovery 2 Millions of "weak symbol/visibility" messages 4 Build design for v0.26.1 4.8 Support for MinGW 5 Refactoring the Tiff Code 5.1 Background 5.2 How does Exiv2 decode the ExifData in a JPEG? 5.3 How is metadata organized in Exiv2 5.4 Where are the tags defined? 5.5 How do the MakerNotes get decoded? 5.6 How do the encoders work? 6 Using external XMP SDK via Conan ========================================================================== 4 Build design for v0.26.1 Added : 2017-08-18 Modified: 2017-08-23 The purpose of the v0.26.1 is to release bug fixes and experimental new features which may become defaults with v0.27 4.8 Support for MinGW MinGW msys/1.0 was deprecated when v0.26 was released. No support for MinGW msys/1.0 will be provided. It's very likely that the MinGW msys/1.0 will build. I will not provide any user support for MinGW msys/1.0 in future. MinGW msys/2.0 might be supported as "experimental" in Exiv2 v0.26.2 ========================================================================== 5 Refactoring the Tiff Code Added : 2017-09-24 Modified: 2017-09-24 5.1 Background Tiff parsing is the root code of a metadata engine. The Tiff parsing code in Exiv2 is very difficult to understand and has major architectural shortcomings: 1) It requires the Tiff file to be totally in memory 2) It cannot handle BigTiff 3) The parser doesn't know the source of the in memory tiff image 4) It uses memory mapping on the tiff file - if the network connection is lost, horrible things happen - it requires a lot of VM to map the complete file - BigTiff file can be 100GB+ - The memory mapping causes problems with Virus Detection software on Windows 5) The parser cannot deal with multi-page tiff files 6) It requires the total file to be in contiguous memory and defeats 'webready'. The Tiff parsing code in Exiv2 is ingenious. It's also very robust. It works well. It can: 1) Handle 32-bit Tiff and Many Raw formats (which are derived from Tiff) 2) It can read and write Manufacturer's MakerNotes which are (mostly) in Tiff format 3) It probably has other great features that I haven't discovered - because the code is so hard to understand, I can't simply browse and read it. 4) It separates file navigation from data analysis. The code in image::printStructure was originally written to understand "what is a tiff?" It has problems: 1) It was intended to be a single threaded debugging function and has security issues. 2) It doesn't handle BigTiff 3) It's messy. It's reading and processing metadata simultaneously. The aim of this project is to 1) Reconsider the Tiff Code. 2) Keep everything good in the code and address known deficiencies 3) Establish a Team Exiv2 "Tiff Expert" who knows the code intimately. 5.2 How does Exiv2 decode the ExifData in a JPEG? You can get my test file from http://clanmills.com/Stonehenge.jpg 808 rmills@rmillsmbp:~/gnu/github/exiv2/exiv2/build $ exiv2 -pS ~/Stonehenge.jpg STRUCTURE OF JPEG FILE: /Users/rmills/Stonehenge.jpg address | marker | length | data 0 | 0xffd8 SOI 2 | 0xffe1 APP1 | 15288 | Exif..II*...................... 15292 | 0xffe1 APP1 | 2610 | http://ns.adobe.com/xap/1.0/.accept(reader); The class tree for the decoder is: class TiffDecoder : public TiffFinder { class TiffReader , class TiffFinder : public TiffVisitor { class TiffVisitor { public: //! Events for the stop/go flag. See setGo(). enum GoEvent { geTraverse = 0, geKnownMakernote = 1 }; void setGo(GoEvent event, bool go); virtual void visitEntry(TiffEntry* object) =0; virtual void visitDataEntry(TiffDataEntry* object) =0; virtual void visitImageEntry(TiffImageEntry* object) =0; virtual void visitSizeEntry(TiffSizeEntry* object) =0; virtual void visitDirectory(TiffDirectory* object) =0; virtual void visitSubIfd(TiffSubIfd* object) =0; virtual void visitMnEntry(TiffMnEntry* object) =0; virtual void visitIfdMakernote(TiffIfdMakernote* object) =0; virtual void visitIfdMakernoteEnd(TiffIfdMakernote* object); virtual void visitBinaryArray(TiffBinaryArray* object) =0; virtual void visitBinaryArrayEnd(TiffBinaryArray* object); //! Operation to perform for an element of a binary array virtual void visitBinaryElement(TiffBinaryElement* object) =0; //! Check if stop flag for \em event is clear, return true if it's clear. bool go(GoEvent event) const; } } } The reader works by stepping along the Tiff directory and calls the visitor's "callbacks" as it reads. There are 2000 lines of code in tiffcomposite.cpp and, to be honest, I don't know what most of it does! Set a breakpoint in src/exif.cpp#571. That’s where he adds the key/value to the exifData vector. Exactly how did he get here? That’s a puzzle. void ExifData::add(const ExifKey& key, const Value* pValue) { add(Exifdatum(key, pValue)); } 5.3 How is metadata organized in Exiv2 section.group.tag section: Exif | IPTC | Xmp group: Photo | Image | MakerNote | Nikon3 .... tag: YResolution etc ... 820 rmills@rmillsmbp:~/gnu/github/exiv2/exiv2/src $ exiv2 -pa ~/Stonehenge.jpg | cut -d' ' -f 1 | cut -d. -f 1 | sort | uniq Exif Iptc Xmp 821 rmills@rmillsmbp:~/gnu/github/exiv2/exiv2/src $ exiv2 -pa --grep Exif ~/Stonehenge.jpg | cut -d'.' -f 2 | sort | uniq GPSInfo Image Iop MakerNote Nikon3 NikonAf2 NikonCb2b NikonFi NikonIi NikonLd3 NikonMe NikonPc NikonVr NikonWt Photo Thumbnail 822 rmills@rmillsmbp:~/gnu/github/exiv2/exiv2/src $ 533 rmills@rmillsmbp:~/Downloads $ exiv2 -pa --grep Exif ~/Stonehenge.jpg | cut -d'.' -f 3 | cut -d' ' -f 1 | sort | uniq AFAperture AFAreaHeight AFAreaMode ... XResolution YCbCrPositioning YResolution 534 rmills@rmillsmbp:~/Downloads $ 823 rmills@rmillsmbp:~/gnu/github/exiv2/exiv2/src $ The data in IFD0 of is Exiv2.Image: 826 rmills@rmillsmbp:~/gnu/github/exiv2/exiv2/src $ exiv2 -pR ~/Stonehenge.jpg | head -20 STRUCTURE OF JPEG FILE: /Users/rmills/Stonehenge.jpg address | marker | length | data 0 | 0xffd8 SOI 2 | 0xffe1 APP1 | 15288 | Exif..II*...................... STRUCTURE OF TIFF FILE (II): MemIo address | tag | type | count | offset | value 10 | 0x010f Make | ASCII | 18 | 146 | NIKON CORPORATION 22 | 0x0110 Model | ASCII | 12 | 164 | NIKON D5300 34 | 0x0112 Orientation | SHORT | 1 | | 1 46 | 0x011a XResolution | RATIONAL | 1 | 176 | 300/1 58 | 0x011b YResolution | RATIONAL | 1 | 184 | 300/1 70 | 0x0128 ResolutionUnit | SHORT | 1 | | 2 82 | 0x0131 Software | ASCII | 10 | 192 | Ver.1.00 94 | 0x0132 DateTime | ASCII | 20 | 202 | 2015:07:16 20:25:28 106 | 0x0213 YCbCrPositioning | SHORT | 1 | | 1 118 | 0x8769 ExifTag | LONG | 1 | | 222 STRUCTURE OF TIFF FILE (II): MemIo address | tag | type | count | offset | value 224 | 0x829a ExposureTime | RATIONAL | 1 | 732 | 10/4000 236 | 0x829d FNumber | RATIONAL | 1 | 740 | 100/10 827 rmills@rmillsmbp:~/gnu/github/exiv2/exiv2/src $ exiv2 -pa --grep Image ~/Stonehenge.jpg Exif.Image.Make Ascii 18 NIKON CORPORATION Exif.Image.Model Ascii 12 NIKON D5300 Exif.Image.Orientation Short 1 top, left Exif.Image.XResolution Rational 1 300 Exif.Image.YResolution Rational 1 300 Exif.Image.ResolutionUnit Short 1 inch Exif.Image.Software Ascii 10 Ver.1.00 Exif.Image.DateTime Ascii 20 2015:07:16 20:25:28 Exif.Image.YCbCrPositioning Short 1 Centered Exif.Image.ExifTag Long 1 222 Exif.Nikon3.ImageBoundary Short 4 0 0 6000 4000 Exif.Nikon3.ImageDataSize Long 1 6173648 Exif.NikonAf2.AFImageWidth Short 1 0 Exif.NikonAf2.AFImageHeight Short 1 0 Exif.Photo.ImageUniqueID Ascii 33 090caaf2c085f3e102513b24750041aa Exif.Image.GPSTag Long 1 4060 828 rmills@rmillsmbp:~/gnu/github/exiv2/exiv2/src $ The data in IFD1 is Exiv2.Photo The data in the MakerNote is another embedded TIFF (which more embedded tiffs) 829 rmills@rmillsmbp:~/gnu/github/exiv2/exiv2/src $ exiv2 -pa --grep MakerNote ~/Stonehenge.jpg Exif.Photo.MakerNote Undefined 3152 (Binary value suppressed) Exif.MakerNote.Offset Long 1 914 Exif.MakerNote.ByteOrder Ascii 3 II 830 rmills@rmillsmbp:~/gnu/github/exiv2/exiv2/src $ The MakerNote decodes them into: Exif.Nikon1, Exiv2.NikonAf2 and so on. I don't know exactly it achieves this. However it means that tag-numbers can be reused in different IFDs. Tag 0x0016 = Nikon GPSSpeed and can mean something different elsewhere. 5.4 Where are the tags defined? There's an array of "TagInfo" data structures in each of the makernote decoders. These define the tag (a number) and the tag name, the groupID (eg canonId) and the default type. There's also a callback to print the value of the tag. This does the "interpretation" that is performed by the -pt in the exiv2 command-line program. TagInfo(0x4001, "ColorData", N_("Color Data"), N_("Color data"), canonId, makerTags, unsignedShort, -1, printValue), 5.5 How do the MakerNotes get decoded? I don't know. It has something to do with this code in tiffcomposite.cpp#936 TiffMnEntry::doAccept(TiffVisitor& visitor) { ... } Most makernotes are TiffStructures. So the TiffXXX classes are invoked recursively to decode the maker note. #0 0x000000010058b4b0 in Exiv2::Internal::TiffDirectory::doAccept(Exiv2::Internal::TiffVisitor&) at /Users/rmills/gnu/github/exiv2/exiv2/src/tiffcomposite.cpp:916 This function iterated the array of entries #1 0x000000010058b3c6 in Exiv2::Internal::TiffComponent::accept(Exiv2::Internal::TiffVisitor&) at /Users/rmills/gnu/github/exiv2/exiv2/src/tiffcomposite.cpp:891 #2 0x00000001005b5357 in Exiv2::Internal::TiffParserWorker::parse(unsigned char const*, unsigned int, unsigned int, Exiv2::Internal::TiffHeaderBase*) at /Users/rmills/gnu/github/exiv2/exiv2/src/tiffimage.cpp:2006 This function creates an array of TiffEntries #3 0x00000001005a2a60 in Exiv2::Internal::TiffParserWorker::decode(Exiv2::ExifData&, Exiv2::IptcData&, Exiv2::XmpData&, unsigned char const*, unsigned int, unsigned int, void (Exiv2::Internal::TiffDecoder::* (*)(std::__1::basic_string, std::__1::allocator > const&, unsigned int, Exiv2::Internal::IfdId))(Exiv2::Internal::TiffEntryBase const*), Exiv2::Internal::TiffHeaderBase*) at /Users/rmills/gnu/github/exiv2/exiv2/src/tiffimage.cpp:1900 #4 0x00000001005a1ae9 in Exiv2::TiffParser::decode(Exiv2::ExifData&, Exiv2::IptcData&, Exiv2::XmpData&, unsigned char const*, unsigned int) at /Users/rmills/gnu/github/exiv2/exiv2/src/tiffimage.cpp:260 #5 0x000000010044d956 in Exiv2::ExifParser::decode(Exiv2::ExifData&, unsigned char const*, unsigned int) at /Users/rmills/gnu/github/exiv2/exiv2/src/exif.cpp:625 #6 0x0000000100498fd7 in Exiv2::JpegBase::readMetadata() at /Users/rmills/gnu/github/exiv2/exiv2/src/jpgimage.cpp:386 #7 0x000000010000bc59 in Action::Print::printList() at /Users/rmills/gnu/github/exiv2/exiv2/src/actions.cpp:530 #8 0x0000000100005835 in Action::Print::run(std::__1::basic_string, std::__1::allocator > const&) at /Users/rmills/gnu/github/exiv2/exiv2/src/actions.cpp:245 5.6 How do the encoders work? I understand writeMetadata() and will document that soon. I still have to study how the TiffVisitor writes metadata. 6 Using external XMP SDK via Conan Section 1 describes how to compile the newer versions of XMP SDK with a bash script. This approach had few limitations: 1) We had to include sources from other projects into the Exiv2 repository: Check the folder xmpsdk/third-party. 2) Different scripts for compiling XMP SDK on Linux, Mac OSX and Windows. 3) Lot of configuration/compilation issues depending on the system configuration. Taking into account that during the last months we have done a big effort in migrating the manipulation of 3rd party dependencies to Conan, we have decided to do the same here. A conan recipe has been written for XmpSdk at: https://github.com/piponazo/conan-xmpsdk And the recipe and package binaries can be found in the piponazo's bintray repository: https://bintray.com/piponazo/piponazo This conan recipe provides a custom CMake finder that will be used by our CMake code to properly find XMP SDK in the conan cache and then be able to use the CMake variables: ${XMPSDK_LIBRARY} and ${XMPSDK_INCLUDE_DIR}. These are the steps you will need to follow to configure the project with the external XMP support: # Add the conan-piponazo remote to your conan configuration (only once) conan remote add conan-piponazo https://api.bintray.com/conan/piponazo/piponazo mkdir build && cd build # Run conan to bring the dependencies. Note that the XMPSDK is not enabled by default and you will # need to enable the xmp option to bring it. conan install .. --options xmp=True # Configure the project with support for the external XMP version. Disable the normal XMP version cmake -DCMAKE_BUILD_TYPE=Release -DEXIV2_ENABLE_XMP=OFF -DEXIV2_ENABLE_EXTERNAL_XMP=ON -DBUILD_SHARED_LIBS=ON .. Note that the usage of the newer versions of XMP is experimental and it was included in Exiv2 because few users has requested it.