There are a few situations in which we might need to trigger a scan manually to properly track inodes, like a directory from an untracked path being renamed into a tracked one. In this case, we could trigger a scan of just the renamed directory, however, we have no easy way to do that and need to do a full scan instead.
The reason this is not currently possible is that our scan depends on glob expanding all the patterns that are configured to be monitored, then iterating over those. This situation can probably be improved if we instead manually scanned the paths and used a globset for validating each subpath we find, this will require some additional effort to find the best way to implement this (we obviously don't want to manually scan the entire filesystem, but we can probably apply some logic to the patterns to find common ancestors for each and scan from the longest prefix match of each set of those).
There are a few situations in which we might need to trigger a scan manually to properly track inodes, like a directory from an untracked path being renamed into a tracked one. In this case, we could trigger a scan of just the renamed directory, however, we have no easy way to do that and need to do a full scan instead.
The reason this is not currently possible is that our scan depends on
globexpanding all the patterns that are configured to be monitored, then iterating over those. This situation can probably be improved if we instead manually scanned the paths and used aglobsetfor validating each subpath we find, this will require some additional effort to find the best way to implement this (we obviously don't want to manually scan the entire filesystem, but we can probably apply some logic to the patterns to find common ancestors for each and scan from the longest prefix match of each set of those).