Inhibit system sleep while torrents are moving #18783
Conversation
Consider a moving torrent unfinished - prevents system suspension inhibition lapsing before torrents are considered 'complete' from a UI perspective.
I think any "moving files" is a valid reason for system sleep to be inhibited. |
Corrected boolean logic, since isFinished() being true was sufficient to return false regardless of isMoving().
Fair enough. I've updated the commit to correct the fix - the first attempt did not actually resolve the issue (my fault for not reading the boolean logic properly). |
But your changes make it to behave inconsistently, don't they? Moving files will affect it only when |
|
Now that i think about it... doesn't the OS know files are moved? Doesn't it prevent sleep during that time? |
IIRC, it does not matter what tasks are performed by certain processes, but only whether the system is used interactively or not. If the system is not used interactively for a while, then this is a reason to switch to sleep mode. |
Correct. It makes sense, given that in a modern OS environment file IO is constantly occurring (although it can certainly be unintuitive, such as when you initiate a large file transfer in shell and come back to find your PC asleep).
When I made this PR, my goal was just to make the minimal changes required to avoid a specific scenario (having sleep inhibited during downloads, choosing a completion action such as shutdown, and the action not succeeding) since I felt that was the most likely situation where this would cause a problem. However, I do agree that a more consistent behaviour change would be better. I would tend towards preventing sleep during move operations entirely, rather than adding another option, but that's my personal preference. I'm happy to work on implementing either solution depending on your thoughts. |
👍 |
Sleep prevention timer now always runs; suspends sleep during torrent move operations.
Style correction Co-authored-by: thalieht <thalieht@users.noreply.github.com>
Restored indentation Co-authored-by: thalieht <thalieht@users.noreply.github.com>
glassez
changed the title
|
@Sentox6 |
While a torrent is considered "finished" upon download completion, a session's handleTorrentFinished() event is only queued after any relevant move event is completed. So if a user chooses a post-download action such as system shutdown, torrents will finish, move, and then shutdown occurs.
However, inhibit system sleep behaviour is managed by an independent timer than checks the native status of torrents. This leads to a particular scenario where the user may expect sleep to be inhibited until downloads complete, followed by the system hibernating/shutting down/etc. But since sleep inhibition lapses as soon as torrents are downloaded, a move operation of sufficient duration may allow the system to suspend, meaning the expected post-download action is not executed.
I'm not sure this solution is ideal - for one, it may lead to a user-initiated location change inhibiting sleep unexpectedly. However, any other solutions would seem to require some architectural changes to how updatePowerManagement() works (i.e. transitioning from a timer to being event-driven) or a rethink of what constitutes "finished". Feedback appreciated.