I think I need to do some more testing and then probably update the existing bug report. The problem I have now with that bug report is that I no longer understand exactly why existing ICS placeholder tags in the file conflict with my attempt trying to run version detection on these files again. This is all getting fairly complex...I would certainly report this in Mantis. Basically, you are saying that the ICS reading ignores the version placeholders altogether, right? (Undesired side effect: having both ICS reading and writing on also obliterates them from files upon syncing.)
Yes, that is exactly the problem for which I have not yet figured out what is going on.If you now run the version detection, do you still get inconsistent results? Does that happen even if you import the images with all the discussed options disabled, including the ICS reading?
"Apparently" if there are ICS tags already present in the files, but not the database and these tags are not read to the database, then trying to run version detection on these files causes some sort of hickup. Essentially it seems like these ICS tags are for some reason not being read to the database at all (no matter how ICS is configured!).
The last thing I tried was:
(1) copying the existing database
(2) deleting all images from that copy
(3) deleting the custom placeholders in preferences
(4) import new folder previously not imported with ICS read on, autosync off, import versioning off
I thought that perhaps the existing, but empty placeholder in the database were the problem, i.e., that for some reason PSU would not recognize the placeholder tags in the files being the same as the ones newly created in the database.
Consequently, I would have expected that the ICS schema being read would create both the versions and again create the placeholders. This did not happen. The versions were created, the placeholder tag in the files, however, was ignored, even if I after the import I selected again all files and read the metadata from the files via CTRL+ALT+S.
No matter what I do. Reading the ICS from the files correctly assigns the JPG as main, the RAW files as subversions, but in no scenario ever is the placeholder tag being read from the file!
Now, I have one final attempt left: Creating a completely new, empty database, import all labels again using Hert's script and then import the files with auto-sync off, ICS reading on, import versioning off, but no custom placeholders created in preferences. Perhaps there is a conflict still (even despite having completely deleted the custom placeholders) in my existing database. Perhaps the conflicting placeholders are still there somewhere behind the scenes.
My "logic" behind this: Theoretically I would assume that importing a file with a placeholder tag in the ICS, reading ICS from the file would create that placeholder in the database if it was not previously present. So, if the placeholders were, however, created before in PSU and if PSU for some reason does not recognize that its own placeholders are identical with the tags, then perhaps this is what causes the conflict.
I'll report back.