* Skip scanning zip contents when fingerprint is unchanged
When a zip-based gallery's modification time changes but its content
hash (oshash/md5) remains the same, skip walking and rescanning every
file inside the zip. This avoids expensive per-file fingerprint
recalculation when zip metadata changes without actual content changes.
Closes#6512
* Log a debug message when skipping a zip scan due to unchanged
fingerprint
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: WithoutPants <53250216+WithoutPants@users.noreply.github.com>
* Replace panic with warning if creating a folder hierarchy where parent is equal to current
* Clean stash paths so that comparison works correctly when creating folder hierarchies
* Add root paths parameter to GetOrCreateFolderHierarchy
Ensures that folders are only created up to the root library paths.
* Create full folder hierarchy when scanning a new folder
During a recursive scan, folders should be created as they are encountered (folders are handled in a single thread). This change applies only during a selective scan. Creates up to the root library folder.
* Create folder hierarchy on new file scan
This should only apply when scanning a specific file, as parent folders should be been created during a recursive scan.
* Fix existing folders with missing parents during scan
Adds time-based activity tracking for scenes played via DLNA, enabling
play count, play duration, and resume time tracking similar to the
web frontend.
Key features:
- Uses existing 'trackActivity' UI setting (no new config needed)
- Time-based tracking (elapsed session time / video duration)
- 5-minute session timeout to handle aggressive client buffering
- Minimum thresholds before saving (1% watched or 5 seconds)
- Respects minimumPlayPercent setting for play count increment
Implementation:
- New ActivityTracker in internal/dlna/activity.go
- Session management with automatic expiration
- Integration via DLNA service initialization
Limitations:
- Cannot detect actual playback position (only elapsed time)
- Cannot detect seeking or pause state
- Designed for upstream compatibility (no complex dependencies)
This was originally done for #3304. The ffmpeg code has been redone since and this is no longer necessary. It was also resulting in the scraper and plugin paths being absolute, despite all the others being relative to the provided config path.
* Use more neutral language for content
* Add sfw mode setting
* Make configuration context mandatory
* Add sfw class when sfw mode active
* Hide nsfw performer fields in sfw mode
* Hide nsfw sort options
* Hide nsfw filter/sort options in sfw mode
* Replace o-count with like counter in sfw mode
* Use sfw label for o-counter filter in sfw mode
* Use likes instead of o-count in sfw mode in other places
* Rename sfw mode to sfw content mode
* Use sfw image for default performers in sfw mode
* Document SFW content mode
* Add SFW mode setting to setup
* Clarify README
* Change wording of sfw mode description
* Handle configuration loading error correctly
* Hide age in performer cards
* Move stashbox package under pkg
* Remove StashBox from method names
* Add fingerprint conversion methods to Fingerprint
Refactor Fingerprints methods
* Make FindSceneByFingerprints accept fingerprints not scene ids
* Refactor SubmitSceneDraft to not require readers
* Have SubmitFingerprints accept scenes
Remove SceneReader dependency
* Move ScrapedScene to models package
* Move ScrapedImage into models package
* Move ScrapedGallery into models package
* Move Scene relationship matching out of stashbox package
This is now expected to be done in the client code
* Remove TagFinder dependency from stashbox.Client
* Make stashbox scene find full hierarchy of studios
* Move studio resolution into separate method
* Move studio matching out of stashbox package
This is now client code responsibility
* Move performer matching out of FindPerformerByID and FindPerformerByName
* Refactor performer querying logic and remove unused stashbox models
Renames FindStashBoxPerformersByPerformerNames to QueryPerformers and accepts names instead of performer ids
* Refactor SubmitPerformerDraft to not load relationships
This will be the responsibility of the calling code
* Remove repository references
* Indicate while backing up database
* Close migrate connection to db before optimising
* Don't vacuum post-migration
In most cases is probably not needed and can be an optonal user-initiated step
* Ensure connection close on NewMigrator error
* Perform post-migration using migrator connection
Flush WAL file at end of migration
