Subtracting 1 from the first chapter to download index caused an additional chapter to get downloaded (e.g. limit 4 would download 5 chapters)
Regression was introduced with 05bf4f5525
* Return latest data for finished downloads
In case a download has finished, the cache of the data loader has to be cleared to be able to get the latest data, otherwise, the returned chapter will still be marked as not downloaded
* Correctly clear manga data loader caches
For finished downloads the immediate emission did not work because the emission was done async and by the time the state got updated with the new status, the finished download was already removed from the queue.
Thus, the new state was missing the finished download.
* Remove "updateAvailable" from webui update info
Doesn't add anything
* Extract status creation into function
* Optionally emit status immediately
Otherwise, some emissions can get lost due to the 1s sample period
* Rename "STOPPED" state to "IDLE"
* Reset webui update status
Currently, the update status never gets reset.
Thus, it will be "FINISHED" or "ERROR" until the next update gets triggered.
Due to this, the client won't know that the update result was already handled and will handle it again the next time it gets the update status.
To prevent this, the client has to be able to tell the server that it has handled the update result and that it can be resetted
Due to not immediately sending the status, the finished chapters were already removed from the queue by the time the status was actually send to the client.
This caused the client to never receive a status with the chapters downloaded flag to be true, resulting in the client to not know that a chapter is downloaded
* Use a new type for the webui about info query
Using the same type for this and the webui update queries/mutations caused apollo to save it as the same data in the cache, overwriting the "about info"
* Use a new type for the webui about check query
To prevent similar issues (cc3bf5f34a8afebadd306d037db1a10088ef9334) with the "update check" and the "update progress" payloads
* Throw update check error when calling it via the query
Otherwise, the error is never raised to the frontend
* Set "ERROR" state in case the update check failed on WebUI update trigger
The served file gets cached and thus, it won't reflect the latest version of the file.
This was a problem after the webUI got updated, since now the served "index.html" was outdated and pointed to files that didn't exist anymore.
When adding commits or switching between branches the "shadowJar" gradle task always used the same (outdated) commit count which created jars with confusing names
* Trigger initial auto backup in case server was not running
In case the server was not started (stopped, system shutdown - not in hibernation) during the scheduled auto backup time, the auto backup never got triggered.
* Update server util preferences
* Fix automatic chapter download for initial chapter list fetch
The initial fetch wasn't correctly detected which caused chapters to get downloaded.
Using index based numbers also caused the first chapter to not get downloaded due to it being omitted in the "subList" call which excludes the "toIndex".
* [Logging] Update logs
In case a chapter is marked as not downloaded, but the download folder exists already, the chapter did not get downloaded again.
This could cause issues in case the previous download failed or has missing pages.
Instead of only checking if the folder exists, each page should be checked individually
This was previously done and was incorrectly changed with 1c9a139006.
* Chapter fetch improvements
* Update previous date uploads
* Lint
* Fix backup inserts
* Remove extra maxSeenUploadDate
* Port downloaded over
* Make sure to set isDownloaded on all inserts
The function incorrectly exited early in case no latest read chapter was found.
This rendered disabling the setting "excludeEntryWithUnreadChapters" useless.
* Keep initial fetch date of existing chapters on a list fetch
The fetch at date should not get updated for existing chapters.
Updating this field makes it impossible to detect which chapters were actually newly fetched.
To get the last fetched timestamp of the chapters, the "chaptersLastFetchedAt" field of the manga should be used
got changed in 6d33d72663
* Get real chapter url safely
In case this causes an exception, it should not cause the whole list fetch to fail
was removed in 6d33d72663
Currently, the server tries to migrate the preference on every startup, even if the migration was already done.
This can lead to an unhandled exception, if the write permission to the system preference was revoked.
In case the migration has already happened, these permissions should not be required
* Reset backup status to idle in case of an exception
* Rename "performRestore" function
* Set backup status to failure on exception
Makes it possible to detect if the restore failed or not after the first status was received
* Set backup status to success on completion
Since the status is not provided over a subscription, but over a query that should be pulled, it is not really easily detectable if a restore finished or not, since both states will be indicated by "idle"
* Correctly wait for first new status when triggering backup import
The status is only "Idle" in case no backup import has ever run.
Once the first backup process finished it is either "Failure" or "Success"
* Rename "ProtoBackupImport::restore" function
* Add id to restore process
Makes it possible to differentiate between backup restore processes.
We do not allow any category to have the default categories name ("default" case-insensitive).
In case the backup includes a category with this name, it won't be matched to any category in our database and also won't insert a new category.
This will then lead to an exception, causing the backup to fail.
* Move running check to update function
* Move updating update status to process function
* Fail all source updates in case of update channel failure
In case the channel failed due to an exception, the update for the source failed completely.
This however was never handled and the pending updates for the source were never set to failed.
Due to this, the global updates running state was always true
* Remove completed update channel from available channels
* Always log specific update job failure
* Optionally fetch mangas during the update
The update only fetched the chapter list of a manga but never the manga itself.
Thus, e.g. unless the manga got online fetched via the ui, it would never get recognized if it is completed or not.
This would e.g. prevent the update setting, to not update completed mangas, from working as intended
* Make settings required
Mangas are not supposed to be mapped to the default category in the database.
In case this happens, the category query won't be able to correctly select mangas in the default category
In case the database got deleted without deleting cached/downloaded thumbnails, the next time a manga gets inserted, it's possible that a thumbnail was already downloaded for its id.
This then causes mangas to be displayed with incorrect thumbnails due to using the outdated cached/downloaded thumbnails
In case download ahead is disabled, all new chapters should get downloaded.
Due to incorrectly calculating the index of the first new chapter to download, no new chapter was downloaded at all
In case a manga was already loaded via the data loader, the cached data will get used.
Due to this, the update status did not return the updated manga data, but instead, stale data
* Set default log level to INFO
Default log level was accidentally changed to ERROR for the base logger in 56deea9fb3
* Reduce graphql log level to WARN
Otherwise, thrown exceptions are swallowed by graphql and never logged besides a very brief error in the graphql response
