IsHealAssistantOfIndex() and IsRangedDpsAssistantOfIndex() are supposed
to iterate through the group and first return members with the
applicable role that have the assistant flag, and then iterate through
non-assistants only if there are not enough assistants for the
designated index. They are not written properly and actually completely
ignore the assistant flag.
I rely on these functions for significant roles in SSC and TK (which I
have decided I'll PR in the same way as SSC, as a long-term draft). I
have them fixed on my own fork, but it is problematic for testers if
these functions do not work.
So I've done three things here:
1. Fixed the functions to prefer members with the assistant flag.
2. Added a third parameter for ignoreDeadPlayers, like
IsAssistTankOfIndex() has. Note that the parameter is by default false
for IsAssistTankOfIndex(), meaning dead players are _not_ ignored. This
is not my preferred design choice--I think the default should be to
ignore dead players, but I have not changed the default and have made
the default the same for IsAssistHealOfIndex() and
IsAssistRangedDpsOfIndex(), since I don't know the intent of the
pre-existing boss strats that use the functions.
3. Changed the names to IsAssistHealOfIndex() and
IsAssistRangedDpsOfIndex() so they parallel IsAssistTankOfIndex(), and
made corresponding changes in the few boss strats that use the
functions.
Also, note that the functions _do _not_ exclude real players. I think
there are arguments for and against excluding real players. A fourth
parameter for this could be useful, but I've not made any change in that
regard.
Needs second pair of eyes, they appear in crash logs here and there. Its
merely a patch on a open wound.
----
As in aslong there multithreads in mapupdate, which we need for decent
performance and core calls are not done correctly due various reasons.
These type of issues remain.
Although i am planning to experiment a little with threadsafe execution
of our strategies vs performance.
The most effective thing we could do is check every single action and
check its stateless and where it does effect the state or read the state
of a core object its done in the safest way. flags, worldthread where
possible and/ot simply taking into account the state might be invalid.
Refactor the flightmastercache the bots use for finding the nearest
available flight master.
I made a small change to how the original function worked by storing the
database position for the flightmaster in the cache itself. This allows
us to calculate the distance from bot before accessing the creature
object, should be faster overall.
This adds the bear form as an alternative form for druids with the cat
strategies. This applies to low level (<20) druids that have the cat
strategy applied. Reset botAI would not normally give them the strat,
because they dont have cat form, but if set, this allows more dps.
- Did some basic formatting
- Some generated docs
- Cleaned header/impl Helper.css
- Moved PerfMonitor from util to bot/handler/command
Still a freaking mess though, but its a start i guess. Cant ask ppl to
add more or make use of those when its so messy.
Adding a guard against bots that may be registered in `playerBots`
variable as selfbot have appeared to crash here.
Edit: The exact circumstances that caused the incorrect registration are
under investigation as it seems to be 'atypical' behavior.
**Original issue:**
Bots/Characters received duplicate spells during randomization process.
**Root cause:**
When `PlayerbotFactory::Randomize` is processed, `InitSkills` is called
AFTER `InitAvailableSpells`, which causes the duplicate spells issue
because the skillline ability spell is already learned when processing
spells from trainers (`InitAvailableSpells`).
We simply need to change the order of the randomization process: skills
should be handled first, then spells. An alternative approach would be
to adjust the skillline abilities and check each spell for every skill,
but that seems redundant since we already have checks for the trainer's
spells.
`InitSkills` -> `SetRandomSkill` -> `SetSkill` ->
`learnSkillRewardedSpells` -> `learnSpell` -> duplicate error...
**Steps to reproduce:**
1. create common character and login with it
2. set level for this character eq. 80 (`.set level 79`)
3. create and run macro:
```
/g .playerbots bot initself
/g .saveall
```
4. logout -> login and run macro again
**Note:**
Also added checks for the trainer's spells since `GetSpell` can return
nullptr. Updated `LearnQuestSpells` after recent changes and used the
same logic and implementation from `Player::learnQuestRewardedSpells`.
Yes, we need to cast spells, or we should handle all spell effects that
quests/trainers have (for ex.: `SPELL_EFFECT_SKILL_STEP`,
`SPELL_EFFECT_UNLEARN_SPECIALIZATION`, `SPELL_EFFECT_BIND`).
Replace deprecated vsprintf with vsnprintf to eliminate compiler warning
and prevent potential buffer overflow. Updated to latest commit.
Tested in game and it seemed to log actions just fine. I just basically
added a buffer size by using the current vsnprintf lib instead.
**Problem**: FishingAction::isUseful() used strict `pos ==
bot->GetPosition()` comparison causing "Rog A:go fishing - USELESS" when
bot was within 0.5-1m of fishing spot (normal positioning tolerance).
**Solution**: Replace exact coordinate match with 1.5m tolerance using
`fabs(posX - botX) < 1.5f` for all axes. Added `#include "Config.h"` for
AI_VALUE stability.
**Result**: Master fishing now reliably triggers and completes cycle:
- move near water → go fishing → use bobber ✓
Closes #fishing-useless-bug
---------
Co-authored-by: Кунгуров Олег <okungurov@rapidsoft.ru>
Added two test enchants to exclude from the list of enchants that bots
can apply: 19927 and 44119. I also reordered the list to be in
alphanumeric order.
The important exclusion to add is 44119:
https://www.wowhead.com/wotlk/spell=44119/enchant-bracer-template
It is a test enchant that gives 34 AP on bracers; this is stronger than
any other bracer enchant for level 70 physical attackers so bots are
applying it.
Clarified that MinEnchantingBotLevel also determines whether maintenance
will socket gems, in addition to applying enchants.
Fixed max iLevel for ZA gear in RandomGearScoreLimit description (iLevel
138 is the max that drops from the final boss, but the 3rd timed chest
gives iLevel 141 equipment).
There are two related PVP components in this PR. First is the simple yet
fundamental change to bot behaviour when they are in party. Right now
bots with a master will go into PVP when there's a nearby PVP target,
even if master is not in PVP. This absolutely should not happen. Bots
should not consider PVP at all if master is not in PVP. The fix is only
3 lines in EnemyPlayerValue
The second component is introducing PVP probabilities, to make decisions
more realistic. Right now even a level 1 bot will 100% go into PVP if it
sees a level 80 PVP target. They can't help themselves. So the change
here addresses that insanity. Several thresholds (subject to community
review) are introduced:
1. Bots will not fight a target 5 or more levels higher than them
2. Bots have a 25% chance starting a fight with a target +/- 4 levels
from them.
3. Bots have a 50% chance starting a fight with a target +/- 3 levels
from them.
4. Bots have a 75% chance starting a fight with a target +/- 2 levels
from them.
5. Bots have a 100% chance starting a fight with a target +/- 1 level
from them.
6. Bots have a 25% chance starting a fight with a target 5 or more
levels below them (ganking. thought it would be funny, and technically
realistic of player behaviour)
Exception of course exist for BG/Arena/Duel, and in capitals where bots
will always PVP. Also bots will always defend themselves if attacked.
Few notes:
1. The if/ else if logic can be further simplified, but only if we use
thresholds that are different by one. So current logic allows for
flexibility of using values like 10/7/5/3 instead of 5/4/3/2.
2. The caching system is per-bot basis. So for some target X, if some
bot decides to attack it, another bot will make its own decision. At
first I used a simplified global system (thinking there might be
performance concerns) where if one bot decides to attack a target then
they all do, but when I switched to the more realistic per-bot basis, I
didn't see an effect on performance.
3. Variables are obviously not configurable right now. I'm starting to
see Bash's POV that maybe we have too many configs 😬 Still,
they can be easily exposed in the future, and if someone is reading this
then, remember to change constexpr to const.
Fix for the error "Attempted sending heartbeat with root flag for guid"
The core does not allow movement flag updates when unit/player has the
MOVEMENTFLAG_ROOT flag. Change scope bots alone.
### Summary
- Add a preserveAuras flag to summon teleport helpers.
- Keep auras when using the summon command.
- Leave meeting stone behavior unchanged (auras can still be cleared
there).
### Motivation
Summon command was clearing buffs (e.g., Hellscream’s Warsong in ICC)
because auras were explicitly interrupted before teleporting. This
change keeps existing meeting stone behavior but preserves auras for the
summon path to avoid losing valid buffs.
### Details
Solve :
https://github.com/mod-playerbots/mod-playerbots/issues/1862https://github.com/mod-playerbots/mod-playerbots/issues/1942
### Testing
Go to ICC and téléport Bots
Fix the [issue
#1964](https://github.com/mod-playerbots/mod-playerbots/issues/1964)
where bots on vehicles cant move. Transfer the allowed movement logic to
`PlayerbotAI::CanMove` -- Ulduar raid strategy also used this method for
the Malady of the Mind trigger.
The only remaining `IsRooted` check is in
`MovementAction::UpdateMovementState`. I'm not sure if this also needs
to be updated for the vehicles case, but on paper, everything should
work as intended. A more complex solution to fix this issue can be found
in the
[comment](https://github.com/mod-playerbots/mod-playerbots/issues/1902#issuecomment-3703795779),
but for now, these changes are safer.
<!--
> [!WARNING]
> **This is a DRAFT PR.**
> The structure is not definitive. The code might not be optimised yet.
It might not even start nor compile yet.
> **Don't panic. ✋ It's going to be ok. 👌 We can make modifications, we
can fix things.** 😁
-->
# Description
This PR aims to refactor the NextAction declaration to achieve two
goals:
## Eliminate C-style sentinel arrays
Currently, a double pointer (`NextAction**`) approach is being used.
This an old pre-C++11 (< 2011) trick before `std::vector<>` became a
thing.
This approach is painful for developers because they constantly need to
declare their `NextAction` arrays as:
```cpp
NextAction::array(0, new NextAction("foo", 1.0f), nullptr)
```
Instead of:
```cpp
{ new NextAction("foo", 1.0f) }
```
The first argument of `NextAction::array` is actually a hack. It is used
to have a named argument so `va_args` can find the remaining arguments.
It is set to 0 everywhere but in fact does nothing. This is very
confusing to people unfamiliar with this antiquated syntax.
The last argument `nullptr` is what we call a sentinel. It's a `nullptr`
because `va_args` is looking for a `nullptr` to stop iterating. It's
also a hack and also leads to confusion.
## Eliminate unnecessary pointers for `NextAction`
Pointers can be used for several reasons, to cite a few:
- Indicate strong, absolute identity.
- Provide strong but transferable ownership (unlike references).
- When a null value is acceptable (`nullptr`).
- When copy is expensive.
`NextAction` meets none of these criteria:
- It has no identity because it is purely behavioural.
- It is never owned by anything as it is passed around and never fetched
from a registry.
- The only situations where it can be `nullptr` are errors that should
in fact throw an `std::invalid_argument` instead.
- They are extremely small objects that embark a single `std::string`
and a single `float`.
Pointers should be avoided when not strictly necessary because they can
quickly lead to undefined behaviour due to unhandled `nullptr`
situations. They also make the syntax heavier due to the necessity to
constantly check for `nullptr`. Finally, they aren't even good for
performance in that situation because shifting a pointer so many times
is likely more expensive than copying such a trivial object.
# End goal
The end goal is to declare `NextAction` arrays this way:
```cpp
{ NextAction("foo", 1.0f) }
```
> [!NOTE]
> Additional note: `NextAction` is nothing but a hacky proxy to an
`Action` constructor. This should eventually be reworked to use handles
instead of strings. This would make copying `NextAction` even cheaper
and remove the need for the extremely heavy stringly typed current
approach. Stringly typed entities are a known anti-pattern so we need to
move on from those.
This needs extensive testing.
What's important is spells given to bots. Class spells, mounts,
professions etc. Make sure they get the spells they should, when they
should.
Requires https://github.com/mod-playerbots/azerothcore-wotlk/pull/132
The idea is to centralize the creation, assignment, and management of
bot guilds into a single class that can be referenced.
The way this is intended to work.
when the manager is created, if the config option to delete guilds is
set, then it deletes all bot guilds.
On startup
1. Load all guild names from database. Shuffle keys for some
randomization.
2. Load Guilds from database
3. For existing guilds, identify the guild faction, number of members,
and assess if the guild is 'full' based on the number of bots set in
config.
4. Determine if the leader of the guild is a real player based on the
leader account.
5. Mark any playerbot guild names as not available (false).
The validation process (2-5) is set to run once an hour.
Guild Creation.
Now guild creation occurs on an as needed bases during the
initialization process. Previously, all of the guilds would be created
at once, and then randomly assigned.
When a bot is not in a guild during initialization, it will check if
there are any partially filled guilds of that bots faction where the bot
can be assigned to. If not, and the cache of bot guilds is less than the
set number in config, it will randomly return the available name. This
then goes to the CreateGuild function where the core guild manager
creates a guild, the guild emblem is set, and the cache updated.
If a bot is assigned to guild, but fails to join then it throws an
error.
Checking for real player guilds function now lives in the guild manager.
---------
Co-authored-by: bashermens <31279994+hermensbas@users.noreply.github.com>
Merging this isn't required but rather to provide an option to core
changes.
The scripts can be written from scratch when someone feels like doing it
without touching the core.
This shouldn't be merged unless
https://github.com/mod-playerbots/azerothcore-wotlk/pull/130 is merged
too.
Lines 481 through 484 are the only substantive changes. Bots were having
the potion strategy added in arenas, and since potions are not allowed
in arenas, this caused them to lock up and attempt to repeatedly drink
potions when under the applicable health or mana trigger thresholds. Now
they won't have the potion strategy in arenas.
Otherwise, I noticed a bunch of magic numbers for spec tabs and so went
ahead and did some refactoring:
1. All references to spec tab numbers now use the appropriate constant
name.
2. A bunch of extra braces were deleted.
3. DEATHKNIGHT_TAB was changed to DEATH_KNIGHT_TAB, and
HUNTER_TAB_BEASTMASTERY was changed to HUNTER_TAB_BEAST_MASTERY, to
reflect the correct names of the class and spec.
4. Deleted some comments that were clearly unneeded. There's much more
that can be cleaned up, and probably the entire logic sequence for
adding/removing strategies in AiFactory.cpp can be redone, but that's
above my pay grade.
Introduces PlayerbotsSecureLoginServerScript to handle secure login
scenarios for playerbots. Ensures that if a playerbot is already online
when a login is attempted, it is properly logged out before allowing the
new session. Registers the new script in the Playerbots initialization.
I found an existing config option that allows bots to speak without a
master, and extended its use so that protected commands (nc, co) are
taken from a bot by GM players. This is primarily a debugging tool and
shouldnt be on generally.
My testing of this is the following.
With the option off, no error is seen, but no text output is shown.
With the option on
GM accounts (Whether GM is on or off) will respond to nc and co commands
by just saying the output directly. Strategies can be changed by GMs for
bots whom they are not masters for (Note, im pretty sure that a GM could
change the strategies even before this change, you just couldnt see the
output of that.)
For non GM characters you get "Invite me to your group first" with the
option on/off.
### Update :Thank you to @notOrrytrout from prompting me to work on
this. Its been a huge learning experience.
With @notOrrytrout I started working on enabling bot fishing with
master, but also on their own.
The first commit didnt crash, showing that it was possible to have a bot
cast when master does. Currently it compiles but crashes when you try to
fish with a bot in the group, whether the bot has fishing or not. It
makes me think that the check in FishingValues is broken somehow, but I
cant figure out how.
---------
Co-authored-by: bash <31279994+hermensbas@users.noreply.github.com>
Disable bot XP gain when master has XP turned off feature has a big
performance impact.
Instead of reverting placed back to original code before the feature
PR's for now as quickfix
**Original issue:**
https://github.com/mod-playerbots/mod-playerbots/issues/1902
**Original cause:** charmed bot (with lost client control) got rooted at
the same time.
**How to reproduce:**
1. Spawn creatures 11350 (x3) and 11830 (x3) using the command: `.npc
add <entry>` in a quiet place.
2. Create any party with random bots or alt bots (should be 60-65
levels), make sure there is at least one healer.
3. Set the healer's mana to a high value, like 100M, using command:
`.mod mana <value>`.
4. Start the fight while continuously respawning creatures with: `.resp
all`.
5. When console starts to spam 'heartbeat' messages, check your party
members' movement flags to identify which one has `MOVEMENTFLAG_ROOT`
0x00000800 (2048) using the command: `.debug move`.
This PR will not fix ALL 'heartbeat' issues, as
`ServerFacade::SetFacingTo` still sends `SendMovementFlagUpdate` while
bots can have the `MOVEMENTFLAG_ROOT` flag.
These contains various fixes, fixes that have history worked one in past
more then once as person as group, aswell @Wishmaster117. But due
various reasons we had to drop them due priority or simply timewise.
These fixes have recollected again by @Regrad based on his crash logs.
Most crash logs we have, i am talking 30+ of them, to many to post here.
@Regrad running a larger server 100+ real players with bots, which means
he will walk into issues that most of us wont or are extremely difficult
to reproduce.
@Regrad used LLM to solve them based on crash log and mentioned his
server crashes almost disappeared, instead of redoing every single PR
and pull them apart. I tried to keep his bunch of changes together as
whole, reviewed them, some redone, verified again etc etc. This is not
how would normally do this. But since i want @Regrad being able to
confirm, we need this in a package as a whole. Pulling them apart in the
current situation is simply to much, to complicated in the verification
process.
So this PR is open and in my opinion has priority above others, but
@Regrad is only person who can give the green light for the
mod-playerbot changes for now.
I, we spend huge amount of time into these issues over last couple of
months. I will put other PR's on hold for abit.
---------
Signed-off-by: Engardium <regradius@gmail.com>
Co-authored-by: Engardium <regradius@gmail.com>
This PR fixes#1833 where bots could keep chasing a raid target icon
(usually skull) across very large distances or even different maps after
the mark was set. It replaces the previous attempt with a simpler design
that keeps values generic and moves context logic into triggers/actions.
Changes
- RtiTargetValue now ignores RTI targets that are farther than
sPlayerbotAIConfig->sightDistance from the bot or the master (same map
only), while still using AttackersValue::IsValidTarget and LOS checks.
- AttackersValue is reverted to its original behavior (no special RTI +
IsInCombat logic).
- RTI triggers only react to "rti target" in combat, as suggested in
review (“that condition should be check in trigger…”).
- AttackRtiTargetAction keeps a small local fallback: if "rti target" is
null, it resolves the raid icon directly from the group so chat commands
like “attack rti target” still work, including out of combat.
Behavior
- Bots no longer run across the world to chase a stale skull far away
from the group.
- When the group comes back near the marked mob (within sight distance),
the skull is again used as a normal focus hint.
- Automatic RTI behavior is limited to combat via triggers, while
explicit chat commands still work out of combat.
Testing
- Marked a mob with skull, killed it, moved far away / changed zone,
then let it respawn: bots did not chase it from afar.
- In a normal dungeon/raid pack, bots still focus skull in combat.
- Passing near a previously marked mob: bots only react once it is back
within sight distance.
---------
Co-authored-by: Keleborn <22352763+Celandriel@users.noreply.github.com>
Same deal as Karazhan--changing map id to instance id for timer keys to
prevent conflicts from multiple groups (plus other tweaks with respect
to the timers). This strategy could use a broader refactor, but that's
for another day--this PR is just to address the timer logic.
I'm not 100% sure but think that the current implementation of timers
for Attumen, Netherspite, and Nightbane with the instance's map id as
the key would result in the timers being shared across separate raids
running the instance at the same time. I've refactored to use the
instance id as the key for all shared timers so to the extent this was
an issue, it shouldn't be anymore.
I also made some minor tweaks to the tank positioning for BBW and
Curator as I noticed there was some weirdness in the logic that I
neglected to address with the refactor.
The feature `Disable bot XP gain when master has XP turned off` (#1910)
was unintentionally disabled in #1929.
This PR re-enables it with better checks in order to prevent crashes
when playing together with the individual progression module.
