New event verbose command#9024
Conversation
|
Part of what I had described was logging to be able to see event class fields before and after each listener for an event. That way if an an event had 50 listeners, you could try and find what listener changed what. EDIT: Another thing, is I feel some events should be excluded by default. Like the tick events, ones that are just spammed a ton. Can have something that does include those, but most of the time they are just gonna clog up the system. We can also exclude events who don't have "changeable" values. |
|
Most QoL way to do that would be with an EventExecutor and listen for a specific event - not sure if there's a way to register one internally though. My thought process was that this was just for all of the "what event runs when xyz" questions pop up. Will exclude tick start and end events, not sure what you mean by "changeable" values though. |
|
In the SimplePluginManager#fireEvent and PaperEventManager#callEvent, can't you just get the values of the event before and after each call to the listeners? "changeable" is hard to define. Generally I think events that have all final fields with no mutable things exposed don't need to be included. These events are just events used to inform that something happened, and a listener can't actually change anything about the event. The |
Note: I had to do this in a skewed way cause my git is losing its mind, if this broke something my fault
|
Now ignores all final fields and can take specific exclusions, right now just tick starting and ending Worth keeping this as its own system? Can just make a new command for watching specific events, can also just nuke this feature entirely for what you were more thinking of |
|
I would exclude final collections (Collection, Map) and ItemStack from the final check. It's possible to have a final one of those that can be mutated via a getter. I suppose both are useful. Seeing what events fire between two points, and a deeper look at what changes for each listener for those events. Might be best for the system I imagined to specify a single event to do that for. |
|
Added in the general premise for what you were looking for through Also moved the event verbose to paper's event manager to ignore bad event calls. The final check of that also now excludes |
|
All should be done unless I missed anything |
Owen1212055
left a comment
Owen1212055
left a comment
There was a problem hiding this comment.
Generally, look at your field names, some of them are UPPERCASE_EXAMPLE when not needed.
| + }; | ||
| + } | ||
| + | ||
| + private void dumpToFile(CommandSender sender, String event, List<Wrapper> wrappers) throws IOException { |
There was a problem hiding this comment.
In general, events can quickly accumulate. I think it may be better to instead actively log to a file async, similar to the latest.log that is auto generated. Have it log an event instead of having to store all of these event calls in memory.
There was a problem hiding this comment.
Is latest.log not just sl4j? Would appreciate a point towards an async internal task method
kennytv
added
the
pre-softspoon: never rebased
7 participants
Saw a message from Machine Maker on Discord bringing this up, figured it'd be a fun random thing to do
Essentially just adds a /paper dumpevents command. Running this once will begin logging events, and running it again will save the events in order with timestamp and event class fields in JSON to a file in the debug directory. Attached below is an example output
Two issues that aren't too big of a deal with this
Does not handle inheritance of event classes well, Field#getDeclaredFields only retrieves minimal fields, could use a superclass check working all the way up to org.bukkit.Event
Could further serialize non primitive event fields just for a bit more info, stringified versions of them are typically fine though
event-verbose-2023-03-21_18.03.40.txt