Changed how data pushed to lua is validated to use weak_ptr

[Shauren]

Makes use of TrinityCore/TrinityCore@a79b42b

Also removed additional indirection when pushing 64 bit integers and ObjectGuids

This allows storing direct object references in scripts

local test_stashed_player = nil

local function OnTextEmote(event, player, textEmote, emoteNum, guid)
    if test_stashed_player ~= nil then
        xpcall(
            function() test_stashed_player:SendBroadcastMessage("reuse stored player") end,
            function(err) player:SendBroadcastMessage(err) end
        )
    else 
        test_stashed_player = player
        test_stashed_player:SendBroadcastMessage("store player ".. tostring(player:GetGUID()))
    end
end

RegisterPlayerEvent(24, OnTextEmote)

Player handles are intentionally invalidated on cross map teleports

1. login
2. /lol
3. /lol
4. .tele dalaran
5. .rec
6. /lol
   Result:
   

PR is marked as draft because map create/destroy hooks currently don't work (fix needed in main TC repo that changes where they are called)

Other emus won't currently compile - I plan to submit a PR to CMangos with the same commit that was pushed to TC (in the worst case that nobody is interested in adopting that code it would have to be instead done by forks that maintain Eluna integration)

[Rochet2]

We may still want to have the old system to invalidate pushed objects at end of an event stack.
Or at some specific event, like at start of map updates starting.

One of the issues is that if you have an aura for example aura = player:GetAura(spellid), then that aura will be valid until the aura object is destroyed.
But as I understand, aura can be transferred to other maps with a player.

If we dont use some specific event to always invalidate objects, then it can happen that an aura or other object is transferred to another map and can be accessed from two lua states at the same time during map update events when we run multi-state mode eluna.

Any thoughts on this?

[Shauren]

Actually, you don't see this in the PR but aura handle is not invalidated when its destructor runs (which is delayed a bit, first it gets added to m_removedAuras list) but immediately when it is unapplied from unit (see TrinityCore/TrinityCore@a79b42b#diff-f3fb604c220161af7953902aecc9586825b8193413ab94fdeea4de5665d7a393R622)

But you are right about moving to another map - I will need to think about it since I was planning to use this outside of lua too

[Shauren]

Luckily aura already has another reference it can additionally check, its WorldObject owner - so this is what I came up with e676a56

[Rochet2]

There are also Loot functions proposed here.
I wonder how cases like this should be handled where there is a new type exposed that does not necessarily have the unique_trackable_ptr.

I guess introducing new type would always require modifying the cores first to introduce the tracking and then the type can be exposed after.

[Shauren]

I'm starting to think its would be better to only allow preserving references for longer for some types, not all of them (using the old callbackid for the ones we want to exclude)

[Foereaper]

Currently the primary objects where storing the GUID and fetching the object is the norm is; players, creatures (units) and gameobjects. I'm guessing we can start with those at the very least and expand where it makes sense.

[Shauren]

Should be addressed by 39eed65

[Foereaper]

While working on something else I got reminded of this PR. Thoughts on wrapping the proposed changes in TC ifdefs to preserve cmangos functionality for now, and then remove the legacy code if/when a similar implementation is added to the other emulators?

[Shauren]

Untested: https://github.com/Shauren/mangos-wotlk/commits/unique_trackable_ptr/

[Foereaper]

I'll see if I can get a cmangos server set up to test some time this weekend, unless someone beats me to it

edit:
Pushed a PR (Shauren/mangos-wotlk#1) to fix the missing GetWeakPtr functions. cmangos also does not have support for ToItem or ToWorldObject, so that would need to be added.