[CS:S] PreEvent player_death calling after death

[Jake]

Syntax: Select all

@PreEvent('player_death')
def pre_player_death(game_event) : 
    player = Player.from_userid(game_event['userid'])
    if player.dead : 
        print('player dead already')

I'm trying to destroy a player's weapons before death but the weapons are dropped before this PreEvent is fired. Is there another way?

[Ayuto]

What exactly do you want to achieve with this? One thing I can think of is hooking drop_weapon and check if the dropping player is dead.

[Jake]

I want to remove any weapons that are dropped manually or on death to prevent spam. drop_weapon is what I was looking for, thanks

Syntax: Select all

@EntityPreHook(EntityCondition.is_player, 'drop_weapon')
def pre_drop_weapon(args) : 
    if args[1] : 
        weapon = make_object(Weapon, args[1])
        weapon.remove()

[L'In20Cible]

Since you do nothing else on the Weapon object, you could optimize quite a bit:

Syntax: Select all

# Can't indent first line? hm...
    weapon = make_object(BaseEntity, args[1])
    weapon.destroy()

[BackRaw]

Since you do nothing else on the Weapon object, you could optimize quite a bit:

Syntax: Select all

# Can't indent first line? hm...
    weapon = make_object(BaseEntity, args[1])
    weapon.destroy()

A little bit OT but: how is destroy() different from remove()?

[L'In20Cible]

The result will be the same, but execution time will be way faster. remove() is bound to the Kill input (so we internally have to loop through all classes, lookup the datamaps, etc. till we find the correct pointer to dynamically call) while destroy() is wrapping its factory (direct call to what the input itself is doing internally - minus the fact that we have to internally look for it in the global dictionary), plus the fact that you skip two classes in the hierarchy to instanciate (Weapon → Entity → BaseEntity), in that specific case. However, remove_entity() would have been faster, but got removed (not sure why) which is also a regression since now we internally rely on the entity's classname instead of only its pointer (which breaks the update of my parachute plugin (where the model is literally an entity named "parachute") now that I think about it).

[Ayuto]

However, remove_entity() would have been faster, but got removed (not sure why) which is also a regression since now we internally rely on the entity's classname instead of only its pointer (which breaks the update of my parachute plugin (where the model is literally an entity named "parachute") now that I think about it).

remove_entity() was only working with indexes, so I decided to add the destroy() method. I didn't adopt the implementation of remove_entity(), because it was implemented differently for every engine (I think we even used the factory method in L4D2 and I'm not sure if the hammerid implementation was really working properly). However, I just took a closer look at the implementation of the factory's Destroy() method and it's just calling INetworkable::Release(), which simply deletes the entity immediately. That might be a bit risky. Otherwise I have no idea why the global entity list maintains a delete list. So, the best way might be using UTIL_Remove() or calling the kill input (which calls UTIL_Remove() at some point) in BaseEntity.destroy().

[L'In20Cible]

remove_entity() was only working with indexes, so I decided to add the destroy() method. I didn't adopt the implementation of remove_entity(), because it was implemented differently for every engine (I think we even used the factory method in L4D2 and I'm not sure if the hammerid implementation was really working properly).

Yeah, forgot how messy and hacky it was for some engines. The hammerid patch was working on CS:GO (I think blade was also implemented this way, if I remember correctly), but was feeling pointless.

However, I just took a closer look at the implementation of the factory's Destroy() method and it's just calling INetworkable::Release(), which simply deletes the entity immediately. That might be a bit risky. Otherwise I have no idea why the global entity list maintains a delete list.

I believe the engine is storing deleted entities, so it doesn't reuse the slot before they have been correctly broadcasted to clients. Deleting the entity immediately, may cause issues, according to the documentation of UTIL_Remove:

Syntax: Select all

// Purpose: Sets the entity up for deletion.  Entity will not actually be deleted
//			until the next frame, so there can be no pointer errors.

Which probably makes sure all tasks (motioning, rendering, broadcasting, etc) are done before freeding the actual pointer. Using Destroy()/INetworkable::Release() may have undefined behavior, I guess.

So, the best way might be using UTIL_Remove() or calling the kill input (which calls UTIL_Remove() at some point) in BaseEntity.destroy().

Definetly. In my opinion, we should implement BaseEntity.destroy() as BaseEntity.remove() calling the InputKill function (now that it is possible on the cpp side!) internally. That way we remove the need to rely on the classname, we make sure the entity is marked for deletion and correctly handled over freeding the address and hoping for the best.

A bit off topic but, back to this quote:

remove_entity() was only working with indexes

We currently have a lot of wrapper which accepts index, and internally convert to pointer passing them to engine calls. This was done this way, before BaseEntity, pointers, etc. being wrapped. I guess that now, we should definetly avoid extra cast by making them accepting the pointer directly (and by pointer I mean the actual BaseEntity objects). Some InputVariant/TakeDamageInfo methods, that I can think off. Do we all agree with that? This will break some plugins, but better doing it now.

[Ayuto]

I guess that now, we should definetly avoid extra cast by making them accepting the pointer directly (and by pointer I mean the actual BaseEntity objects). Some InputVariant/TakeDamageInfo methods, that I can think off. Do we all agree with that? This will break some plugins, but better doing it now.

Yep, agreed!

[burnSTATION]

Ideally you would just make this change in your config files

mp_death_drop_gun 0

[Ayuto]

That cvar only exists in CS:GO, but not in CS:S.