Exiv2

I've looked carefully at this. I'm not convinced that there is a bug here. Here's my point of view. I've modified the code (change below) and the output is now as you request:

$ exiv2 -pa -K Exif.CanonCs.FocusContinuous ~gnu/exiv2/trunk/test/tmp/20030925_201850.jpg
Exif.CanonCs.FocusContinuous                 Short       1  (-1)
818 rmills@rmillsmbp:~/gnu/exiv2/trunk $ 

However you will observe that the type of the data in "Short" and not "SShort". The fix requires me to introduce a bug in the code in tags_int.hpp#226

    template <int N, const TagDetails (&array)[N]>
    std::ostream& printTag(std::ostream& os, const Value& value, const ExifData*)
    {
        const TagDetails* td = find(array, value.toLong());
        if (td) {
            os << exvGettext(td->label_);
        }
        else {
            if ( value.typeId() == unsignedShort ) {
                signed short v = (signed short) value.toLong();

                os << "(" << v << ")";
            } else {
                os << "(" << value << ")";
            }
        }
        return os;
    }

So the question is "Why is value being delivered as Short and not an SShort?". The tag is defined as signedShort in canonmn.cpp#1134.

TagInfo(0x0020, "FocusContinuous" 
, N_("Focus Continuous"), N_("Focus continuous setting")
, canonCsId, makerTags, signedShort, 1
, EXV_PRINT_TAG(canonCsFocusContinuous)),

So, I believe there is something in the data in the file itself which overrules the default definition.

I didn't write Exiv2, and there are parts of the code which I have never explored. The decoding of MakerNotes is one of those black holes.

There is a less brutal change that could be made to the code in tags_int.hpp:

const TagDetails* td = find(array, value.toLong());

This code reads the tag definition from canonmn.cpp and therefore can discover that although value.typeId() == 3 == unsignedShort , its default is signedShort and should be converted.

My background is writing compilers and interpreters. I strongly dislike this kind of hack. If the data type is recorded by Canon as unsignedShort, it is not the job of Exiv2 to change it on the fly. I will not submit such a change (nor the hack in this report). I've tried setting the type explicitly and this is not successful.

824 rmills@rmillsmbp:~/gnu/exiv2/trunk $ exiv2 -M"set Exif.CanonCs.FocusContinuous SShort -1" cs.jpg
825 rmills@rmillsmbp:~/gnu/exiv2/trunk $ exiv2 -pa -K Exif.CanonCs.FocusContinuous cs.jpg
Exif.CanonCs.FocusContinuous                 Short       1  (65535)
826 rmills@rmillsmbp:~/gnu/exiv2/trunk $ exiv2 -M"set Exif.CanonCs.FocusContinuous Short -1" cs.jpg
Warning: Exif.CanonCs.FocusContinuous: Failed to read Short value "-1"

The road to resolving this is for me to master the Canon MakerNote decoder and I don't have time to undertake that at the moment. If Andreas wrote that code and can give further insight, that would be very helpful.

I do have another constructive change to make here concerning 8 = Manual. I have submitted that change: r4365

822 rmills@rmillsmbp:~/gnu/exiv2/trunk $ cp /Users/rmills/gnu/exiv2/trunk/test/tmp/20030925_201850.jpg cs.jpg
823 rmills@rmillsmbp:~/gnu/exiv2/trunk $ exiv2 -pa -K Exif.CanonCs.FocusContinuous cs.jpg
Exif.CanonCs.FocusContinuous                 Short       1  (65535)
827 rmills@rmillsmbp:~/gnu/exiv2/trunk $ exiv2 -M"set Exif.CanonCs.FocusContinuous Short 8" cs.jpg
828 rmills@rmillsmbp:~/gnu/exiv2/trunk $ exiv2 -pa -K Exif.CanonCs.FocusContinuous cs.jpg
Exif.CanonCs.FocusContinuous                 Short       1  Manual
829 rmills@rmillsmbp:~/gnu/exiv2/trunk $ exiv2 -pv -K Exif.CanonCs.FocusContinuous cs.jpg
0x0020 CanonCs      FocusContinuous             Short       1  8

I'm not sure I've understood your comment about interpreting -1 as Single. If we don't know the meaning of -1, we shouldn't invent a meaning.

As I am trying to get closed on v0.26, I'm don't have time to investigate the SShort/Short contradiction at this time. I'm therefore marking this as "Resolved".

Thanks for opening issue #1203 to deal with the SShort/Short contradiction. I'll deal with this in v0.27.

Can you confirm that you are happy to value -1/63535 is not translated as "Single".