Exiv2

I've updated the test suite to use the test file provided by Markus. Thank you Markus for providing this.

I'm unhappy about this fix. In an earlier submission, I thought I had this fixed by getting StringValueBase::read to respect the nul as a terminator. Indeed that passed test/bugfixes-test.sh and was submitted. However that fix broke test/conversion.sh and other tests as it prematurely terminated unicode strings.

There is a test in TiffReader::readTiffEntry() for tag Exif.Canon.LensModel and to respect nul. This fix doesn't scale as more instances of this behavior will require special handling.

I've revisited the subject of nul-terminated ascii strings in Canon MakeNotes.

Code was added to tiffvisitor.cpp in January (r2966) to give special treatment to Exif.Canon.LensModel. Although I suspect there are other tags with this behavior, I found no vidence using the files in <exiv2dir>/test/data. For example, the length every string in 01.jpg is correct:

653 rmills@rmills-imac:~/gnu/exiv2/exiv2.master/test/data $ exiv2 -pt 01.jpg | grep Ascii
Exif.Image.Make                              Ascii       6  Canon
Exif.Image.Model                             Ascii      12  Canon EOS M
Exif.Image.DateTime                          Ascii      20  2012:06:29 14:48:14
Exif.Image.Artist                            Ascii       1  
Exif.Image.Copyright                         Ascii       1  
Exif.Photo.DateTimeOriginal                  Ascii      20  2012:06:29 14:48:14
Exif.Photo.DateTimeDigitized                 Ascii      20  2012:06:29 14:48:14
Exif.MakerNote.ByteOrder                     Ascii       3  II
Exif.Canon.ImageType                         Ascii      12  Canon EOS M
Exif.Canon.FirmwareVersion                   Ascii      22  Firmware Version 5.2.3
Exif.Canon.OwnerName                         Ascii      32  
Exif.Canon.LensModel                         Ascii      16  EF-M22mm f/2 STM
Exif.Canon.InternalSerialNumber              Ascii       9  EA0002893
Exif.Photo.SubSecTime                        Ascii       3  88
Exif.Photo.SubSecTimeOriginal                Ascii       3  88
Exif.Photo.SubSecTimeDigitized               Ascii       3  88
Exif.Iop.InteroperabilityIndex               Ascii       4  R98
Exif.Photo.CameraOwnerName                   Ascii       1  
Exif.Photo.BodySerialNumber                  Ascii      13  001009900099
Exif.Photo.LensModel                         Ascii      17  EF-M22mm f/2 STM
Exif.Photo.LensSerialNumber                  Ascii      11  000000002d
654 rmills@rmills-imac:~/gnu/exiv2/exiv2.master/test/data $ 

I've undone the changes in r2966 (and related in r2973 and r2976) and added a fix on the output side instead (exactly what you said in http://dev.exiv2.org/issues/839#note-2, Robin): We read the complete tag value from the image but when output to a stream is requested, only write up to the first \0 character. That's more generic and more inline with the general exiv2 philosophy as it ensures the tag value won't change - even if it contains \0 characters - unless explicitly requested.

Thanks for dealing with this, Andreas.