GNOME Bugzilla – Bug 621066
Change duplicate detection to work on whole-file hash at import.
Last modified: 2010-06-11 12:16:58 UTC
Duplicate detection is currently broken in subtle ways. For this reason I propose to change duplicate to work on the MD5 hash of the entire file. This hash is calculated when the file is imported and will never be recalculated. Advantages: * Faster, no need to generate a thumbnail * Does perfect duplicate detection of files that come from a camera / card Disadvantages: * We need to nuke all the duplicate data * Changed files which are reimported won't get the right import hash
Currently there's a check in place that refuses to import a file if it would get the same name as an existing file. No actual duplicate detection is done in that case. Obviously this is very broken, but I suggest we leave it in place for a couple releases, until people have a sufficient amount of new hashes in their database (which get calculated during import). Comments?
Built and merged to master.