Safe getTileBlock

[dvantwisk]

I'm writing a renderer that has to repeatedly access the getTileBlock() method in Maps.cpp. I've had my debugger show crashes here before at the return statement:

return world->map.block_index[x >> 4][y >> 4][z];

It seems that it is trying to access an invalid memory address even though the coordinates are valid. This indicates sometimes dfhack can hit the map_block data before it has updated, causing the crash. Adding a check for null pointers seems to have resolved the crashes on my end. Let me know if this is a prudent change or if I ought to be doing something else when having to make many calls to getTileBlock().

[quietust]

Under normal circumstances, world->map.block_index, world->map.block_index[x], and world->map.block_index[x][y] should never be NULL as long as x/y are valid, but world->map.block_index[x][y][z] can definitely be NULL if a particular block happens to be "unallocated"*. If you're ever seeing intermediate NULL pointers, it's possible something managed to run while DF was in the process of rebuilding block_index (which I'm not sure how that would happen, unless you were walking around in Adventurer mode).

(* - though such unallocated blocks seem to be exceedingly uncommon in the current version - back in 0.47.05 and earlier, blocks up in the Sky were often unallocated, and way back in 0.28.181.40d and earlier, nearly all underground blocks were unallocated until you placed a designation inside them).

[dvantwisk]

The custom renderer I'm working on makes many calls to getTileBlock (although I'm rewriting a few parts to cut back on it). You're right that the situation that the block is unallocated is rare, however, I'd think the check is still prudent.