Old XMP's are silently discarded, thus breaking any backward compatibility.
Noise profile with dartable-cli of 2.2.0rc0 does produce different histogram than in 2.0.7 (Version of debian stretch), profile can't be calculates with 2.2.0rc0
$\033[31mERROR: Incorrect green channel.
o The input RAW image is bad regarding lighting or exposure.
o You may have set a default output profile in darktable which is
unsuitable for noise profiling.
I always used the tools from 2.2.0rc0 to create the profile
#1 Updated by Roman Lebedev over 3 years ago
- % Done changed from 0 to 10
- Priority changed from Low to Critical
- Assignee set to Tobias Ellinghaus
- Status changed from New to Confirmed
- Subject changed from Noise profiling does not work to Old XMP's are silently discarded, thus breaking any backward compatibility.
#5 Updated by Johannes Hanika over 3 years ago
for what it's worth, with this little patch i can load the old xmp file again.
in short, back when we wrote xmp like this, everything was in <rdf:Bag>, not <rdf:Seq>. very much unsure about the implications of this, i think we'd need to run this function twice. once with the Seq and if that fails retry with the Bag variant? please note the patch is just a demonstration and very much incomplete.
i'm attaching old-broken.xmp for reference and easier reproduction, too.
also i don't think we need roman's revert to solve this. probably this pugi xml code path was always broken for these really ancient xmp files?