Pathfinder::prependCells() crashes the game if cell is null

[roossienb]

Likely caused by the cell->getParentCell() in the for-loop definition on line 9061, which can return NULL

Map and replay.zip

[Mauller]

The funny thing there is they even test for null but do nothing about it.

This probably wants catching earlier in the code if possible to prevent cell being assigned a null pointer.

This loop could even be a single null cell in the list since the game ran out of available path finding cells.

[roossienb]

Update so far:
cell isn't null.
The call to getXIndex() and getYIndex() make use of an unchecked instance m_info which is null.

m_info is set in allocateInfo(). AllocateInfo return if the creation of m_info is successful or not. If not, there are no more paths available.

There are a number of allocateInfo calls that do not check if the allocation was successful

[Mauller]

This issue will likely track down the cause of and resolve the following

• Crash when too many units are moving #330
• Game crashes when a lot of units are being moved at once #81

[Mauller]

Added the generals tag as a quick look shows that the return from allocateinfo is also unhandled in the same places as within zero hour.

[Mauller]

Image with Stack Trace

[roossienb]

This is going to be a difficult one.

• m_map holds cells on the map. It looks like some of these cells can have parents, because they are part of a path,
• When creating a new path, I'm getting a cell from m_map that has a chain of parents, which end up with a parent without m_info,
• The cause is releaseClosedList(),
• There is another chain in the cells through cell->m_nextOpen,
• At some point, the list that thats closed, contains an cell (via m_nextOpen) that also appears in m_map (in a different chain). Because the cell is released, the chain in m_map is truncated and the highest parent is now stuck without m_info.,
• Question is: why can a cell be in a parent-train and in a nextOpen chain at the same time?

[Mauller]

So i believe i have a fix for this and it also resolves #979