[crash] Waypoints

[Artkeep]

9435

../game/libmangosgame.a(CreatureAIRegistry.o): In function `MovementGeneratorMedium<Creature, WaypointMovementGenerator<Creature> >::GetResetPosition(Unit&, float&, float&, float&)':

CreatureAIRegistry.cpp:(.text._ZN23MovementGeneratorMediumI8Creature25WaypointMovementGeneratorIS0_EE16GetResetPositionER4UnitRfS6_S6_[MovementGeneratorMedium<Creature, WaypointMovementGenerator<Creature> >::GetResetPosition(Unit&, float&, float&, float&)]+0x15): undefined reference to `DestinationHolder<Traveller<Creature> >::GetLocationNowNoMicroMovement(float&, float&, float&) const'

collect2: ld returned 1 exit status

make[3]: *** [mangos-worldd] Ошибка 1

make[3]: Leaving directory `/server/Mangos-Sources/compile/compile-9435-/objdir/src/mangosd'

:mad:

[The_Game_Master]

Any ideeas?

[DrKLO]

if we look at code we see

if (creature.IsStopped()){ - look at this check
....
MovementInform(creature);
}

and after

if (i_nextMoveTime.Passed())
   {
       // If stopped then begin a new move segment
       if (creature.IsStopped()) - the same check
       {
        ....
       }
       else
      {
          ....
           i_nextMoveTime.Reset(i_path->at(i_currentNode).delay); - there was crash, because i_past = null;
           ...
       }

if first check passed, crash code can't be call

so making check i_path after MovementInform(creature); as in [9434] doesn't help

looks like it's deleted from other thread, but creatures on what was crash don't have eventAI or SD2 scripts

PS sorry for my bad english

[VladimirMangos]

"looks like it's deleted from other thread" this is impossible in current core. Diff map-threads not supported...

i_path changes only in WaypointMovementGenerator<Creature>::LoadPath and in WaypointMovementGenerator<Creature>::ClearWaypoints()

ClearWaypoints called from ReloadPath not used in clean code now. So or object deleted or LoadPath called somewhere

But in single thread if Update called this possible only after MovementInform... Where deleting/NULL cases checked....

[SmDevil]

DrKLO, I think you need add thread id in Log... It's very helpful debug MT... And it's will arguments for your words... (sorry for English)

@@ -20,10 +20,11 @@
#include "Log.h"
#include "Policies/SingletonImp.h"
#include "Config/ConfigEnv.h"
#include "Util.h"
#include "ByteBuffer.h"
+#include "ace/Thread_Manager.h"

#include <stdarg.h>

INSTANTIATE_SINGLETON_1( Log );

@@ -287,11 +288,11 @@ void Log::outTimestamp(FILE* file)
    //       MM     month (2 digits 01-12)
    //       DD     day (2 digits 01-31)
    //       HH     hour (2 digits 00-23)
    //       MM     minutes (2 digits 00-59)
    //       SS     seconds (2 digits 00-59)
-    fprintf(file,"%-4d-%02d-%02d %02d:%02d:%02d ",aTm->tm_year+1900,aTm->tm_mon+1,aTm->tm_mday,aTm->tm_hour,aTm->tm_min,aTm->tm_sec);
+    fprintf(file,"%-4d-%02d-%02d %02d:%02d:%02d %d ",aTm->tm_year+1900,aTm->tm_mon+1,aTm->tm_mday,aTm->tm_hour,aTm->tm_min,aTm->tm_sec, ACE_Thread::self());
}

void Log::outTime()
{
    time_t t = time(NULL);
@@ -300,11 +301,11 @@ void Log::outTime()
    //       MM     month (2 digits 01-12)
    //       DD     day (2 digits 01-31)
    //       HH     hour (2 digits 00-23)
    //       MM     minutes (2 digits 00-59)
    //       SS     seconds (2 digits 00-59)
-    printf("%02d:%02d:%02d ",aTm->tm_hour,aTm->tm_min,aTm->tm_sec);
+    printf("%02d:%02d:%02d %d ",aTm->tm_hour,aTm->tm_min,aTm->tm_sec, ACE_Thread::self());
}

std::string Log::GetTimestampStr()
{
    time_t t = time(NULL);

PS: And we all understood that current realization of core (single thread) is "die". Have you a plans about it?

[The_Game_Master]

Are you saying that the current problem can't be solved in the current core ST state?

[SmDevil]

I think so, Vladimir said that it's problem can't be in ST. But DrKLO said that SD2 starting in other thread in current realization of core, but it's need check. And patch help him...

[The_Game_Master]

I managed to crash it without SD2 a couple of hours ago. Just plain mangos. I don't think it's SD2 related. And if it is, as vladimir said "fix the script".

[Wojta]

And...when it appears?

I have mangos 9389 and SD2 1573 without these crashes.

It cannot just appear from nowhere....

[VladimirMangos]

I managed to crash it without SD2 a couple of hours ago. Just plain mangos. I don't think it's SD2 related. And if it is, as vladimir said "fix the script".

I not reject core problem case, but need some repeatable way for testing

[DrKLO]

But DrKLO said that SD2 starting in other thread in current realization of core, but it's need check.

i say that it's look like, not sure

Vladimir, can after MovementInform(creature); - creature.IsStopped() become false?

Russian:

Если не может, то краш точно не из за этого, т.к. краш только тогда, когда creature.IsStopped() = false.

[The_Game_Master]

Crash implemented in 9400-9401 if you didn't know until now.

I not reject core problem case, but need some repeatable way for testing

I'm trying to reproduce it for 3 hours now. First time it crashed i forgot to record the method, and now ... it's just pissing me off.

[VladimirMangos]

crash showuo not meaning that bug added in specific changes, it can be just hidden old bug.

[DrKLO]

CreatureTraveller traveller(creature);

i_nextMoveTime.Update(diff);
i_destinationHolder.UpdateTraveller(traveller, diff, false, true); -this

function DestinationHolder<TRAVELLER>::UpdateTraveller
...
traveller.Relocation(x, y, z, ori); - in some cases - this

template<>
inline void Traveller<Creature>::Relocation(float x, float y, float z, float orientation)
{
   i_traveller.GetMap()->CreatureRelocation(&i_traveller, x, y, z, orientation);
}

void
Map::CreatureRelocation(Creature *creature, float x, float y, float z, float ang)
...
// if creature can't be move in new cell/grid (not loaded) move it to repawn cell/grid
// creature coordinates will be updated and notifiers send
if(!CreatureRespawnRelocation(creature)) - this

bool Map::CreatureRespawnRelocation(Creature *c)
...
c->GetMotionMaster()->Clear(); - this

in 9400-9401 there was added some changes in CreatureRelocation, and maybe some of this changes delete WaypointMovementGenerator? so we have crash, and i_path = 0

[DrKLO]

if someone compile source with #define MANGOS_DEBUG and enable logs, we should see what happened before crash in logs

maybe we should change in WaypointMovementGenerator.cpp this

// prevent a crash at empty waypoint path.
   if (!i_path || i_path->empty())
   {
       creature.clearUnitState(UNIT_STAT_ROAMING_MOVE);
       return true;
   }

   // i_path was modified by chat commands for example
   if (i_path->size() != i_hasDone.size())
       i_hasDone.resize(i_path->size());

   if (i_currentNode >= i_path->size())
       i_currentNode = 0;

and this

   CreatureTraveller traveller(creature);

   i_nextMoveTime.Update(diff);
   i_destinationHolder.UpdateTraveller(traveller, diff, false, true);

to avoid crash

[The_Game_Master]

If you make a patch i can give it a quick test. But i don't think that's the source.

[MrGhost]

Hi guys, I just want to ask how frequent your server(s) are crashing. Just to find out if I'm having more issues (or not). Because I see lots of people who still reach 30 mins uptime.

My server often crashes after a few seconds, and around 4 minutes. Reaching 30 minutes is very rare at this moment, this is with around 300+ people while using the latest rev.

[The_Game_Master]

30 secs to 30 minutes uptime.

There are around 12 crash problems in the latest revision. The waypoint related one is the worst.

[MrGhost]

Thanks for your reply, I was thinking I screwed up something (by adding patches).

I also noticed from my crashlogs that these waypoint crashes appear a lot yes.

[Quriq14]

I think the cause may be some use of method MotionMaster::Initialize(), that reloads WaypointMovementGenerator. It is an unsafe function (deletes every MG without the checks of Clear() or MovementExpired()), so should be called only in safe places.

This doesn't happen, for example, here or here:

Map.cpp, line 1026:

   // teleport it to respawn point (like normal respawn if player see)
   if(CreatureCellRelocation(c,resp_cell))
   {
       c->Relocate(resp_x, resp_y, resp_z, resp_o);
       c->GetMotionMaster()->Initialize();                 // prevent possible problems with default move generators
       c->SetNeedNotify();
       return true;
   }
   else
       return false;

There may be other points in code.

[piroy1337]

just revert that movement thingie, till vlad gets an fix for it

[silvermcd]

30 secs to 30 minutes uptime.

There are around 12 crash problems in the latest revision. The waypoint related one is the worst.

there are some crashs too related to the dungeon finder tool

[VladimirMangos]

CreatureTraveller traveller(creature);

i_nextMoveTime.Update(diff);
i_destinationHolder.UpdateTraveller(traveller, diff, false, true); -this

Yes, you find source! I will fix shortly this and other places affected by this.

[silvermcd]

Yes, you find source! I will fix shortly this and other places affected by this.

found the problem?

[VladimirMangos]

In [9445] added checks that must avoud this problem. Thanks to DrKLO for research