Exiv2

Hi Axel

I'm surprised that Andreas hasn't said "thanks for the patch", so I'm guessing that's he's on vacation. I always leave Andreas to deal with patches and I'm sure he'll update and thank you for your contribution shortly.

Thanks for posting this.

Robin

Hi Axel,

Thanks for your patch and sorry for the late reply. I'm not on vacation anymore but still only slowly getting back into the normal rhythm.
Anyway, I've checked-in your patch with r2289, with some modifications and together with a few unrelated fixes.

• The output for Exif.CanonSi.MeasuredEV2 if the value is 0 is now -6.00. Does that make sense?
• What's the unit of these values? EV? (The common practice is to write the unit out)

Regards,
Andreas

To answer your questions:

• The output for Exif.CanonSi.MeasuredEV2 if the value is 0 is now -6.00. Does that make sense?

-> for one thing, those numbers (gain and offset) are actually from ExifTool. Other than that my own empirical tests show, that they are roughly equivalent to the output of MeasuredEV when in evaluative or weighted-average mode. EV2 is independent of the set metering mode and seems to come from a second light meter that seems to average the whole field of view.

• What's the unit of these values? EV? (The common practice is to write the unit out)

-> From my understanding, those values are to be considered unit-less as they are merely used to index a table with equivalent exposure settings (see 'EV as an indicator of camera settings' in http://en.wikipedia.org/wiki/Exposure_value)

Here comes a somewhat embarrassing story... 4 years ago I wrote the above patch and Andreas applied a modified version of it. Unfortunately, his version was broken and I did not notice at the time. I used my own patched binary until the next official release would include the patch. Then I forgot about it and only now thought I could finally replace my old binary with the standard Debian build only to find out that my MeasuredEV feature worked only half the time. After quite some debugging and looking at the wrong places, I found the missing cast to 'signed short'.

Please apply the attached patch against svn HEAD.

Thanks to Axel for reporting and providing the patch. Submitted r3617. I've reported this as Issue: #1037