Your point about #3 does indeed look fishy. I'm not sure why they're calling the destructor on their own anyways.
Try this: // Retire this shake. void *shake = m_ShakeList.Element( nShake ); m_ShakeList.FastRemove( nShake ); delete shake; continue; -----Original Message----- From: hlcoders-boun...@list.valvesoftware.com [mailto:hlcoders-boun...@list.valvesoftware.com] On Behalf Of Maarten De Meyer Sent: Thursday, March 31, 2011 11:38 AM To: Discussion of Half-Life Programming Subject: [hlcoders] CUtlVector crash ( screenshake ) Hi List! So, we're finally out there in closed beta. One of the most annyoing issues is a client side crash. The call stack doesn't say all that much, but still: it always happens in the Purge command in the destructor of CUtlVector: template< typename T, class A > inline CUtlVector<T, A>::~CUtlVector() { Purge(); } more precisely, in the CUtlVector< screenshake_t * > . [ There's only one of those, in CViewEffects ]. Now, some questions: 1) any idea why this happens? [ had to ask :D ] 2) if no, any idea on how to debug? 3) Is the following normal? : // Retire this shake. delete m_ShakeList.Element( nShake ); m_ShakeList.FastRemove( nShake ); continue; where: template< typename T, class A > void CUtlVector<T, A>::FastRemove( int elem ) { Assert( IsValidIndex(elem) ); Destruct( &Element(elem) ); if (m_Size > 0) { memcpy( &Element(elem), &Element(m_Size-1), sizeof(T) ); --m_Size; } } and template <class T> inline void Destruct( T* pMemory ) { pMemory->~T(); #ifdef _DEBUG memset( pMemory, 0xDD, sizeof(T) ); #endif } Concretely: how come it is allowed to first delete a pointer, and afterwards still call the destructor on it? Is that not dangerous? [ Probably a c++ finesse I'm missing ] Shake works fine most of the time though ( I tried with the console cheat command ), so it's not a reproducible crash yet - but happens very often in online games. Thanks in advance for any lights shed in the darkness! -- Maarten _______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlcoders _______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlcoders
