[15:02:04] <BotoX> so there is clearly something rong with sdktools_entoutput
[15:02:18] <BotoX> HookSingleEntityOutput sometimes doesn't fire
[15:02:29] <BotoX> and HookEntityOutput sometimes fires for the wrong thing
[15:08:02] <psychonic> I think at least the former has been known for a while, but no one has been able to work out the cause, largely due to the issue being very intermittent
[15:11:48] <asherkin> I've heard both of those - I thought they were fixed
[15:14:26] <BotoX> running latest master branch so guess not
[15:14:39] <BotoX> and yeah i wasn't able to reproduce in single player
[15:16:11] <asherkin> BotoX: try searching the tracker
[15:18:22] <psychonic> asherkin: i don't remember anything being fixed with those
[15:28:11] <Fyren> Looking at the code history, not much has changed in a long time.
[15:32:26] <Fyren> https://forums.alliedmods.net/showthread.php?t=168696
[15:32:31] <Fyren> I remember nothing about this even though I replied.
[15:33:19] <BotoX> well
[15:33:22] <BotoX> it randomly fires
[15:33:56] <BotoX> or rather
[15:33:59] <BotoX> most of the time it works
[15:34:02] <BotoX> and sometimes it doesn't :D

[19:46:53] <BotoX> I think I'll just get rid of the "fast lookup" in sdktools EntityOutputManager
[19:48:26] <BotoX> it saves the raw address of the entity object in a trie
[19:48:44] <BotoX> but i don't see it deleting it anywhere when that entity dies
[19:49:05] <BotoX> so if that entity dies and another one is created
[19:49:09] <BotoX> it'd have the same address, no?
[19:49:15] <BotoX> as they're all in the same fat array
[19:52:20] <psychonic> yuck
[19:59:24] <Fyren> BotoX: Looking at the code, the trie holds a list of hooks.
[19:59:29] <Fyren> There's no reason to remove things from the trie.
[19:59:39] <Fyren> The list of hooks has references and they do get cleaned up.
[19:59:52] <BotoX> Q_snprintf(sOutput, sizeof(sOutput), "%p", pOutput);
[19:59:57] <BotoX> pOutput == this of entity
[20:00:02] <BotoX> EntityOutputs->Retrieve(sOutput, (void **)&pOutputName)
[20:02:06] <Fyren> "Saved in a trie" is not the same as "key for the trie."
[20:02:25] <BotoX> well a key is also saved in the trie :^^^^^^^^^^^^^^)
[20:02:41] <Fyren> I would not say so, no.
[20:02:58] <BotoX> anyways, that looks wrong, does it?
[20:03:03] <Fyren> It could be.
[20:20:51] <BotoX> are you looking into this Fyren, else I will just rip the "fast lookup" out of there
[20:20:55] <BotoX> throw it on the server
[20:20:58] <BotoX> and if the issues are gone
[20:21:01] <BotoX> report back
[20:24:35] <Fyren> BotoX: I'm kind of looking, but am very unconvinced the lookup matters.
[20:25:12] <BotoX> inserting memory addresses of entities into a trie
[20:25:15] <BotoX> idk 
[20:25:25] <BotoX> i don't trust that at all
[20:25:50] <Fyren> It's not right but I don't think it'd break the hook.
[20:26:23] <Fyren> The addresses are of the output objects, not the actual entities, not that it's really any different.
[20:26:27] <BotoX> but it might be the reason for it firing completly wrong stuff
[20:26:49] <Fyren> It iterates all the saved hooks for the address and checks the ent refs.
[20:27:10] <Fyren> If the saved ent ref doesn't match the ref of what actually called the output, then it removes that hook entry.
[20:27:28] <Fyren> If the entity dies and that address is never reused, it does leak that memory, effectively.
[20:27:31] <BotoX> ah i see yeah
[20:27:50] <BotoX> but then 
[20:27:53] <BotoX> it got the wrong thing
[20:28:00] <BotoX> so it wouldn't fire that one time
[20:28:02] <BotoX> and next round
[20:28:04] <BotoX> it'd fire again
[20:28:10] <BotoX> this is actually kinda the behavior i was seeing
[20:28:36] <Fyren> The next time the detour happens, it'll check the ref.
[20:28:50] <Fyren> If you hooked a single entity, the ref could be different but then that's your fault.
[20:29:01] <BotoX> i hook single entities
[20:29:08] <BotoX> and what would be my fault lol
[20:29:20] <Fyren> If every round your entity gets killed and recreated.
[20:29:32] <BotoX> oh yeah i hook it onentitycreated every round
[20:29:36] <BotoX> right i am stupid
[20:29:48] <BotoX> so this is not the issue
[20:29:54] <Fyren> If you add the hook every round, then the second time around it skips the first old/dead hook.
[20:29:57] <Fyren> The ref won't match.
[20:30:01] <Fyren> Then removes it from the list.
[20:30:18] <Fyren> But it intends to still go through the whole list and should find the newer entry.
[20:31:05] <BotoX> though i think it is getting the wrong OutputNameStruct
[20:31:16] <BotoX> so my hook is not even in the list
[20:33:54] <Fyren> What output do you think it's missing?
[20:34:11] <Fyren> Can you write a script that like creates/hooks/makes the output happen/kills and repeats?
[20:35:08] <BotoX> it's getting the OutputNameStruct with all the hooks of that output
[20:35:15] <BotoX> using the address of the output
[20:35:20] <BotoX> now if that output was free'd
[20:35:32] <BotoX> and a new one allocd on that address
[20:37:27] <BotoX> / same entity index but different reference. Entity has changed, kill the hook.
[20:37:30] <BotoX> in this case
[20:37:36] <Fyren> If a totally different output landed on that address, maybe you could get additional incorrect callbacks.
[20:37:44] <BotoX> which i am also seeing
[20:37:57] <BotoX> HookEntityOutput("env_entity_maker", "OnEntitySpawned", OnEnvEntityMakerEntitySpawned);
[20:38:07] <Fyren> The output name would be wrong so it'd be easy to notice when it did happen.
[20:38:12] <BotoX> getting random other shit calling OnEnvEntityMakerEntitySpawned
[20:38:49] <BotoX> let me grep my logs for that 
[20:39:50] <BotoX> grep "\[SOURCEMOD BUG\] output:" errors_201* | wc -l
[20:39:50] <BotoX> 1299
[20:40:34] <BotoX> errors_20170213.log:L 02/13/2017 - 11:57:17: [BossHP.smx] [SOURCEMOD BUG] output: "OnEntitySpawned", caller: 975, activator: 975, delay: 0.000000, classname: "point_spotlight"
[20:40:34] <BotoX> errors_20170213.log:L 02/13/2017 - 11:58:36: [BossHP.smx] [SOURCEMOD BUG] output: "OnEntitySpawned", caller: 577, activator: 2, delay: 0.000000, classname: "func_breakable"
[20:40:40] <BotoX> i stopped logging it a while ago
[20:41:04] <BotoX> I am pretty sure a point_spotlight or func_breakable can not fire OnEntitySpawned :)
[20:41:18] <BotoX> and they sure as hell are not env_entity_maker either heh
[20:41:43] <Fyren> Are you hooking that globally or single?
[20:41:53] <BotoX> that is global
[20:41:55] <BotoX> <BotoX> HookEntityOutput("env_entity_maker", "OnEntitySpawned", OnEnvEntityMakerEntitySpawned);
[20:42:07] <BotoX> on single i don't think i am getting wrong calls
[20:42:12] <Fyren> I think the extra wrong callbacks would only happen globally.
[20:42:12] <BotoX> well it wouldn't matter since i filter them so idk
[20:42:18] <BotoX> but on single i am missing calls
[20:47:57] <Fyren> Anyway, a way to reproduce is better than trying shots in the dark.
[20:48:35] <Fyren> The extra callbacks could have a low-effort fix by adding the class name and checking it.
[20:49:11] <BotoX> the effort to reproduce is a lot higher
[20:49:17] <BotoX> than just removing the code that looks fishy to me
[20:49:57] <Fyren> Changing code because of "who knows" probably won't motivate anyone to commit it.
[20:50:06] <BotoX> well if it works after that
[20:50:13] <Fyren> Could change the trie to use a string for the key instead with the classname + outputname.
[20:50:13] <BotoX> then i know if that is at fault
[20:50:14] <BotoX> or not
[20:50:32] <BotoX> and look into it more
[20:50:40] <BotoX> without trying to write tests blind
[20:57:59] <Fyren> If you change the key, I'll review and commit.  It'll at least fix the extra callbacks.
[20:58:19] <Fyren> If you just remove the lookup and keep a single, giant list of every output hook, not really.
[20:59:44] <Fyren> But first, the post office.
[21:00:33] <BotoX> well all i did now was completely remove the fast lookup
[22:21:46] <BotoX> so far no issues by removing fast lookup
[22:21:48] <BotoX> fixed™
[22:21:54] <BotoX> not like that shit is needed anyways