Jobs weird on reload?

[embolalia]

5.x reload uses bind_commands, which clears the queue. This (I think) cancels all the jobs, meaning all scheduled stuff probably breaks when any module is reloaded. The 6.0 branch doesn't do anything with jobs, so old versions of jobs keep running. I haven't verified this behavior, but it's probably the case.

[embolalia]

Tweaked so that this is no longer any worse than what previous versions had, so this no longer blocks 6.0 from being released. It's still not right, though, so I'm leaving it open.

[lramati]

Might be a complicated fix, but what if we saved some sort of reference to every job a module sets up so that when the module is unloaded/reloaded we can kill it? add some function to the JobScheduler that takes a Job object (or the ID of said object, or some other unique indicator) and removes all instances of said job from the queue?

[dgw]

Possibly related: #1053, which does away with the clear_jobs() call and loops through the intervals defined on the callable being unregister()ed.

[Exirel]

Will be fixed in #1479.

[Exirel]

It was already fixed in 6.5.x (or was it 6.6.x?, @dgw will know) but not yet is fully available in master.

[dgw]

It was already fixed in 6.5.x (or was it 6.6.x?, @dgw will know) but not yet fully available in master.

Nope, 6.6.x still clears the whole queue:

sopel/sopel/bot.py

Lines 219 to 222 in 684c0b4

	if hasattr(obj, 'interval'):
	# TODO this should somehow find the right job to remove, rather than
	# clearing the entire queue. Issue #831
	self.scheduler.clear_jobs()

Nevertheless, it's fixed on master now with #1479 merged, so… yay.