destroying a terraindeformation

[Narks]

How do I destroy a terraindeformation handle? (as in, remove the handle, so no leaking)

Unrelated note: do I need to null unit handles? (I've heard that they leak anyway)

 

[tooltiperror]

The only function in common.j that takes a [LJASS]terraindeformation[/LJASS] as an argument, to my knowledge, is [LJASS]native TerrainDeformStop[/LJASS]. I'll test if it leaks and edit this post ...

Anything that extends handle needs to be nulled, including widgets, and therefore units.

 

[Romek]

You need to null any handle that could be destroyed. This excludes players, most triggers, and most boolexprs. Units won't directly be removed through triggers, but will die and decay eventually.
Also, you usually don't need to null heroes (unless they're going to be physically removed) or any dummy units that are recycled, since they don't decay and won't be removed. Though you're better of nulling every unit rather than checking if it's a hero or not.

If you don't know what not to null, you're better of nulling everything that extends handle.

 

[PurgeandFire]

I'll test if it leaks and edit this post ...

tooltip broke his promise.

Anyway, [ljass]TerrainDeformStop[/ljass] does remove the leak.

 

[Narks]

awesome, thank you!

 

[WaterKnight]

Anyway, [ljass]TerrainDeformStop[/ljass] does remove the leak.

It does not. Or why do the ids climb up?

function Trig_Nahkampf_Initialisierung_Actions takes nothing returns nothing
    local terraindeformation t = TerrainDeformCrater(0, 0, 64, 100, 1000, true)

    call TerrainDeformStop(t, 100)

    call BJDebugMsg(I2S(GetHandleId(t)))

    set t = null
endfunction

//===========================================================================
function InitTrig_Nahkampf_Initialisierung takes nothing returns nothing
    set gg_trg_Nahkampf_Initialisierung = CreateTrigger(  )
    call TriggerRegisterPlayerEventEndCinematic( gg_trg_Nahkampf_Initialisierung, Player(0) )
    call TriggerAddAction( gg_trg_Nahkampf_Initialisierung, function Trig_Nahkampf_Initialisierung_Actions )
endfunction

Tested this with temporary and permanent deformations, with immediate TerrainDeformStop/fading and with putting a wait between creation and stopping, every combination.

Not every handle needs to be nulled. These with a high id have to be, which are generally agents. Do not ask me why

triggeraction

is no agent. Most non-agent handles are static types like

unitstate

or local ones like

lightning

which have no reference-counter.

 

[Bribe]

handle pointers consume RAM just by existing, and if there are some handles that have no reference counter it would surprise me and I would like to see proof / list.

 

[WaterKnight]

I did not mean that local non-agent handle pointers get released at the end of their function. This, I do not know. Ok, that could be a reason to nullify them if that did help. But said objects are not reference-counted, already freeing their id upon the destroy-function.

function Trig_Nahkampf_Initialisierung_Actions takes nothing returns nothing
    local lightning l = AddLightning("MBUR", false, 0, 0, 0, 0)

    call DestroyLightning(l)

    call AddLightning("LEAS", false, 0, 0, 0, 0)

    call MoveLightning(l, false, 0, 0, 1000, 1000)
endfunction

//===========================================================================
function InitTrig_Nahkampf_Initialisierung takes nothing returns nothing
    set gg_trg_Nahkampf_Initialisierung = CreateTrigger(  )
    call TriggerRegisterPlayerEventEndCinematic( gg_trg_Nahkampf_Initialisierung, Player(0) )
    call TriggerAddAction( gg_trg_Nahkampf_Initialisierung, function Trig_Nahkampf_Initialisierung_Actions )
endfunction

Variable l points to MBUR first, MBUR gets deleted, restores the id although l still on it, LEAS obtains the just returned id and can be accessed through l, drawing the lariat to given coordinates.

As lightnings and some other stuff are not relevant for gameplay and local-only, I guess you do not need as much security.

 

[Bribe]

Then I'll run some tests and figure out if the RAM increases with more unfreed pointers to lightnings, texttags, etc. If they don't hurt RAM, don't affect the handle stack, then there is probably no reason to null them?

 

[PurgeandFire]

>It does not. Or why do the ids climb up?

Yeah, you're right. I accidentally was reading the id of null, so I tested it again and it seems that it does leak. =(

 

[Narks]

:x

So terraindeformations cannot be destroyed?

 

[PurgeandFire]

I'm afraid not, at least from what I see. I tried both of the stop natives, and neither worked to solve it.