* Update inotify event handling where files might be temporarily moved by applications. This scenario is common with certain text editors (like Vim with specific configurations), which can lead to misleading file deletion detections.
* The OneDrive API does not present a hash for all files, most notably those that are zero byte in size (this may be fixed by the OneDrive API in the future). Add a wrapper to the existing makeItem function to test the file size before working out if this is a going to be a problem and if it is, then print out an error message if the file size is greater than 0 including either the full filename path or the items id.
* Remove sha1 from being used by the client as this is being depreciated by Microsoft in July 2023 - https://devblogs.microsoft.com/microsoft365dev/deprecation-of-sha1hash-on-onedrive-personal/
* Complete the removal of crc32 as this is also no longer present for a long time, but some code elements still existed
* Only compute quickXorHash, not quickXorHash and sha256Hash as computing sha256Hash is CPU expensive
* Update cache database stored items to only store quickXorHash and sha256Hash values (remove crc32 and sha1)
* Due to many poor Internet sources of how to use this application, may folk still do not add --synchronize or --monitor switches because their documentation source fails to detail this. When this is the case, the itemdb does not get fully initialised, but when we exit, we try and clean up, because we have partially initalised, we cannot perform a vacuum of the database which generates an error. This patch adds a flag for this specific scenario, so that if triggered, we do not try to vacuum the database and not triggering a false error.
* Update cgi.d to commit 680003a - last upstream change before requiring `core.d` dependency requirement
* Fix deprecation warnings with dmd 2.103.0 by find/replace `isVirtualFunction` with `isVirtualMethod` in cgi.d
* Fix for extra encoded quotation marks surrounding Docker environment variables
---------
Co-authored-by: Hattum van <Nigel.van.Hattum@rabobank.nl>
Co-authored-by: abraunegg <alex.braunegg@gmail.com>
* Update the GitHub version check to also utilise the date a release was done, to allow 1 month grace period before generating obsolete version message.
* When a parental path is to be excluded, and it is an exact match to be excluded, but further 'included' by a wildcard path match, the parental path exclude needs to take precedence, so that empty local directories are not created when they are mean to be excluded
* Update the detection handling of enabling any configuration item change for the first time that required a --resync. Previously initial config changes were being missed.
* When exiting the application when using --synchronize, display that the sync is complete if we are exiting cleanly and all sync operations are complete
* Add a try & catch block for testing if the file exists locally to catch any filesystem error that may be generated
* Test path to be valid if a symbolic link
* Add developer option 'display_processing_time' to control if performance timing is outputted or not
* If option is enabled, print performance data around how long key functions are taking to process data to assist with understanding any performance related questions
* Update skip notification handling for the following scenarios:
* Invalid Name (Microsoft Naming Convention)
* Invalid Name (Contains an invalid whitespace item)
* Invalid Name (Contains HTML ASCII Code)
* Invalid Item (Invalid symbolic link)
* Rather than force exit if unable to lock the database, add a function and boolean to control if the database access has been init was successful. If not, use the exit scopes to exit the application
* Fix that the --resync option is wrongly enforced when the OneDrive client is started with --list-shared-folders option or other non-sync options. This only happens when configuration changes are detected.
* When enabling system logging to a log file, the actual ERROR line is forced to a new line in the application log. The reason for this is the \n prefix in the error message, which was in place so that when performing CLI logging or systemd logging, the error message would be displayed clearly. This change removes the \n from the actual error message, but inserts a newline before the error message is displayed (and also in some cases post error message) - thus keeping the application runtime look and feel, but improving the application log output.
* Change the exit code of the application when a --resync is required from a generic exit code of 1 to exit code 126, as 126 is supposed to be used when a command invoked cannot be executed - which, in this case, when a --resync is needed - the original command cannot be executed.
* This allows|enables watching for a 126 error code in automated environments, so that when 126 as the exit code is seen, the command can be re-tried potentially automatically with the appropriate resync arguments added.
* Add missing logfile output when enabling logging so that when uploaded new & modified files are skipped, this is correctly reflected in logfile output
* Implement --display-running-config to display the running configuration as used at application startup
* Add ONEDRIVE_DISPLAY_CONFIG variable to control --display-running-config in Docker environments
* Only safe remove the files from the local file system if these are exclusively available for this running instance to perform a delete - otherwise this allows multiple clients to perform a --resync on the same data, which, could lead to a data loss scenario
* When the client needs to exit due to an issue, ensure that the curl http instance is shutdown before the exit is performed. This also potentially solves some segmentation faults seen on Ubuntu|Debian platforms due to issues in the shared library libphobos2-ldc-shared.so.X
* contrib: remove bash hashbang from completion
This file starts with the #! sequence that marks interpreted scripts, but
it is a bash completion script that is merely intended to be sourced.
* src: spelling error (Attemtping => Attempting)
* src: spelling error (reponse => response)
* src: spelling error (sucessfully => successfully)
* Update OneDrive API response handling for National Cloud Deployments
* Add developer option to allow easy switch between /children and /delta to query OneDrive for changes
In some OneDrive Business scenarios, the shared folder /delta response lacks the 'root' drive details. When this occurs, this creates the following error: A database statement execution error occurred: foreign key constraint failed. Ensure we query independently the root details for this shared folder and ensure that it is added before we process the /delta response
* Only update the database if the item modified time is different, otherwise we are needlessly updating the database with data that is the same
* Generating a file hash, during the file integrity check is computationally expensive. Only generate a file hash if the modified time of the file is different, otherwise it is pointless to generate the file hash during each integrity check
* Only flush SHM and WAL post integrity check
* Implement a check to validate local filesystem available space before attempting file download
* Implement 'space_reservation' config option with a default value of 50 MB
* Prevent the original run-away logging error message 'Failed initialization on handle XXXX' from occurring if the system is out of space
* Update documentation and man page
* Fix where if the --confdir as specified has incorrect parent permissions, the client is unable to read the required hash files, thus cause an application crash
* When opening files, we should only be opening them as read-only
* When we write out any hash files. they should only be readable|writeable by the userid that is running the application
* Force a synchronization of a specific folder, only when using --synchronize --single-directory and ignoring all non-default skip_dir and skip_file rules
* When using the application with 'defaults' with logging enabled and no verbose mode, the application output that a sync process has started & finished is only written to the logfile every 5 sync times (25 mins). This change updates this so that, if defaults are being used with logging enabled and no verbose mode, the application will always, at a minimum, write out to the application log file when a sync was started and when it was completed. This provides important reference in the log as to when an activity began and was completed.
* In some scenarios users may have a running background service syncing data (systemd or otherwise). In these events, running the application in standalone mode can create a conflict when attempting to perform database queries when there are significant changes to data structures occurring. This PR specifically checks if the database operation files are present, and, if these are present (meaning the application is most likely currently operational) fail fast and not execute the application as a second instance against the same active configuration.
Potentially also resolves RBZ2061430 and RBZ2075468
* When a new folder is created locally, use the lastModifiedDateTime as provided by OneDrive for the folders timestamp instead of the time when the client created the directory
Note: This only will impact new folder that are created, will not touch or modify existing folders
* Update logging output when running in --monitor --verbose mode in regards to the inotify events
* Add additional logging specifically for delete event to denote in log output the source of a deletion event when running in --monitor mode
* Extend GitHub version check to be done once per day when running in --monitor
* Change notification to also notify GUI if notifications are being utilized
* Add application version check, to check GitHub latest release tag and compare to the application version being run. If the application version is older, display a warning message that the client is obsolete and unsupported and that the user needs to upgrade their client version.
* Explicitly set libcurl options to ignore SIGPIPE, use TCP NAGLE and ensure we are re-using connections. These should be 'enabled by default' by libcurl, but we need to ensure we are setting them to their supposed default
Potentially resolves the following issues: #494, #753, #792, #884, #1162, #1408, #1520, #1526 + others.
* Update application logging configuration instructions
* Update error notification handling when unable to write to required logfile due to access permission issues
* Expand on the edge case solution fix for #1586 to drop --single-directory qualifier in this very specific scenario, and add the DB tie record in all --resync --upload-only scenarios where the driveId is actually remote
* This change is to create an argument version of the config operation_timeout.
There are many times where one may need to temporarily increase the timeout for an operation of a single large file that takes longer than the configured timeout. The command line argument is an excellent way to override the default or 'config' set option if required.
* In fixing #1712 which was due to a OneDrive API change, the flagging that quota is being restricted was missed, thus, if this is truly being restricted may cause files to OneDrive Business Shared Folders be unable to be uploaded despite space being available.
