https://bugs.kde.org/show_bug.cgi?id=402154
--- Comment #29 from Joachim Wagner <jwag...@computing.dcu.ie> --- An alternative to relying on UUIDs and sub-volume IDs is to assume mount points of filesystems do not change and to proceed as follows: * Have a persistent table `I` mapping mount points to internal filesystem ID (currently the device number stat.st_dev). * In each run, start with an empty table `M` mapping device numbers to mount points. * During indexing, query stat.st_dev as usual. If stat.st_dev is not yet in `M` find out what the mount point is and add it. Otherwise, obtain the mount point from `M`. (We could do without table `M` but that would be slow and table `M` is expected to stay tiny. If in doubt, use proper cache logic to limit the size of `M`.) * Look up our internal filesystem ID in `I` with the mount point. If not in `I` yet allocate a new ID for it. -- You are receiving this mail because: You are watching all bug changes.
