This is in response to discussion #1859. Due to the nature of this
already terrible handler, I refuse to add any other types of checks for
whether or not this item is banned and lava. If you want to ban lava,
ban a lava bucket. Don't ban an infinite lava bucket, because finite
lava is still infinite lava in this topsy turvy game where the client
rules the world.
This type of change is designed to make it easier to debug problems with
Bouncer's anticheat checks and antigrief checks by allowing server
operators to send better reports. In particular, the logs will help us
diagnose precisely which checks are tripping, and they can cross
correlate the item with the check to adjust the precise check.
More work is needed to add more checks in more places, but you can't
just do this naively and need to actually think when adding things.
It was discovered that LastNetPosition is being checked to see if it's
zero in Bouncer. Then, Bouncer rejects the update. The default is zero
and potentially this can be zero in other ways. The original code in
master (checked at 9f4892f in GetDataHandlers) had an additional write
on LastNetPosition to update it, but this write was not moved over to
Bouncer. Thus, there is a high probability that players are "desync'd"
after LastNetPosition gets stuck at zero and never updates.
These are high priority checks we really want to look at. I want to add
more of these debug statements to all checks in Bouncer and other parts
of GetDataHandlers, but I think this is good enough for now.
Really.. time for bed.
The Any wasn't terrible, but itwould have allow cross sending projectiles within that dictionary. (send gold stab with copper in hand, or vice versa etc)
This change moves the ban checks used to determine, during TileEdit
events, if an item is banned, out of Bouncer and into the newly isolated
ItemBan subsystem. Rather than creating a large pull request for all of
these, I'm just going to create a series of commits and send them in one
at a time. This should result in about one PR per hook that uses item
bans that needs to move.
* Remove commented out warning disable
* Add initial ItemBans segregation infrastructure
* Add shell for initial OnSecondUpdate stuff
* Add comments yo
* Remove duplicated logic
* Split out more item ban code
This part of the fragments work is primarily aimed at reducing the
complexity of OnSecondUpdate in TShock and moving that check out into
the ItemBans subsytem.
Of major note in this is the removal of "check", which was a string
variable that tracked state and replacement of many of the item ban
activities with sane private methods that are at least somewhat
sensible. Obviously there's a lot to be desired in this system and I'm
really going for a run here by trying to continue a branch from so long
ago that I barely even remember the whole point of existence.
Still to do: GetDataHandlers related item ban code needs to be moved
into its own hook in the ItemBan system. Finally, there is a downside to
some of this: we're basically iterating over players again and again if
we keep this pattern up, which is kinda lame for complexity purposes.
* alt j: comment changes
* Move item ban check out of main playerupdate check
Separates out item ban logic from the rest of GetDataHandlers so that
item bans is more isolated in terms of what fragments is asking for.
* alt-j: convert indentation to tabs
* alt-j: fix botching source code
* Move item ban related chest checks out of gdh
* Remove chest item change detection from item bans
It doesn't do anything. If a user removes an item from a chest, it
bypasses this check. If a user adds an item to a chest, the server seems
to persist the change anyway, even if the event is handled. That's a bug
for sure, but fundamentally, it's not the item ban system's fault.
* Revert "Remove chest item change detection from item bans"
This reverts commit 758541ac5c4d4096df2db05ba2a398968113e1e4.
* Fix logic issues related to item ban handling
Re-implements chest item handling and correctly handles events and
returns after setting handled event state.
* Remove TSPlayer.HasProjectilePermission
In infinite wisdom, it turns out this is not a good method for TSPlayer
to have. It just checks the states of things as per what the item ban
system says is banned and then creates implicit relationships to the
projectile ban system.
Doing this effectively knocks down another external reference to the
item ban system outside of the context of the implementation for the
system itself and its related hooks.
This commit also adds context around what the heck is going on with some
of our more interesting checks as per discussions in Telegram with @Ijwu
and @QuiCM.
* Update changelog
* Remove useless ref to Projectile.SetDefaults
* Change item ban to ban based on ID not strings
I think I was so confused as to why we were passing strings everywhere
that I just felt inclined to continue the trend in previous commits.
360 ticks = 6 seconds, according to the wiki, the max possible duration is proportional to dmg taken, so 7.5 seconds would be the max duration (aka 450 ticks)
If there's one thing I learned in this exercise, it's that letting
people add random python scripts to the repo like gpltext.py that
promise to be easier than find and replace is a bad idea.
This renames some methods in the Utils class to better reflect what they
do or what they provide back. It should be clearer what these methods
are for now than before.
Some anti-cheat calls were changed to "You have been Bounced." I don't
think we need to tell clients why they were disconnected so they can
tune their hacks better.
See the changelog entry for this. Basically, it's worse to keep
confirmed broken noclip detection in, even if it detects noclip 90% of
the time, because a random hacker can make a better noclip system and
just evade it 100% of the time by exploiting holes.
Note: even in the previous iteration of the code the search function
returned -1 if it couldn't find it in the system, which would have made
removal of projectiles by ident and owner impossible because they
weren't found anyway.
Either way, this is related to fixing #1584.
