C++ program crashing due to seemingly unrelated array - c++

Question

I'm developing a terrain generator in OpenGL and encounter a strange bug. I can define how long the triangles that the terrain is made from are like this:
#define TRIANGLE_WIDTH 0.25f
#define INVERSE_TRIANGLE_WIDTH 4
With those sizes I create arrays, that are used to store the data of the terrain. I have made some typedefs for the array that look like this:
typedef std::array<std::array<glm::vec3, CHUNK_SIDE_LENGHT>, CHUNK_SIDE_LENGHT> MapDataVec3Type;
typedef std::array<std::array<glm::vec2, CHUNK_SIDE_LENGHT>, CHUNK_SIDE_LENGHT> MapDataVec2Type;
I am using OpenGLs index based rendering. To minimize memory use I automatically calculate if unsigned shorts are sufficient or if I need unsigned ints like this:
#define CHUNK_SIDE_LENGHT (CHUNK_WIDTH * INVERSE_TRIANGLE_WIDTH)
#define CHUNK_SIDE_LENGHT_FLOAT (float)(CHUNK_SIDE_LENGHT)
#define CHUNK_ARRAY_SIZE CHUNK_SIDE_LENGHT * CHUNK_SIDE_LENGHT
#if CHUNK_ARRAY_SIZE >= 65535
#define MAP_INDICES_ARRAY_TYPE unsigned int
#define MAP_INDICES_ARRAY_GL_TYPE GL_UNSIGNED_INT
#else
#define MAP_INDICES_ARRAY_TYPE unsigned short
#define MAP_INDICES_ARRAY_GL_TYPE GL_UNSIGNED_SHORT
#endif
This all works but when the TRIANGLE_WIDTH gets smaller then 0.125 I get a crash when running the program.
I am using Xcode and whenever a program crashes, Xcode shows me where the error is in the assembly code:
SDL-OpenGL-Tests-3`main:
0x100030760 <+0>: pushq %rbp
0x100030761 <+1>: movq %rsp, %rbp
0x100030764 <+4>: subq $0x807f60, %rsp ; imm = 0x807F60
0x10003076b <+11>: movq 0x5b936(%rip), %rax ; (void *)0x00007fff92ccdd40: __stack_chk_guard
0x100030772 <+18>: movq (%rax), %rax
0x100030775 <+21>: movq %rax, -0x8(%rbp)
0x100030779 <+25>: movl $0x0, -0x291c(%rbp)
0x100030783 <+35>: movl %edi, -0x2920(%rbp)
0x100030789 <+41>: movq %rsi, -0x2928(%rbp)
0x100030790 <+48>: leaq 0x4357b(%rip), %rsi ; "/dev/urandom"
0x100030797 <+55>: leaq -0x2948(%rbp), %rax
0x10003079e <+62>: movq %rax, %rdi
-> 0x1000307a1 <+65>: movq %rax, -0x807178(%rbp)
0x1000307a8 <+72>: callq 0x100030090 ; std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >::basic_string<std::nullptr_t> at string:817
0x1000307ad <+77>: leaq -0x2930(%rbp), %rdi
The error is marked by the arrow. From my limited knowledge of assembly I think this part is corespondent to the start of my int main() that looks like this:
int main(int argc, const char * argv[]) {
std::random_device randomDevice;
std::mt19937 randomEngine(randomDevice());
bool capFps = true;
int fpsCap = 60;
double frameTimeCap = double(1e6) / double(fpsCap);
std::chrono::steady_clock::time_point start, end;
std::cout << "Main Thread ID: " << std::this_thread::get_id() << std::endl;
I really don't get how my terrain generation is affecting the random functions of C++ since I don't use the random engine in my terrain generation, I use it to randomly place some light sources.
Update:
Because elsamuko pointed out that I should interchange the C++ random function with some pseudo random number I removed the random functions and now just set all the random values to zero and it still crashes now with this code:
SDL-OpenGL-Tests-3`main:
0x10002fc30 <+0>: pushq %rbp
0x10002fc31 <+1>: movq %rsp, %rbp
0x10002fc34 <+4>: subq $0x807510, %rsp ; imm = 0x807510
0x10002fc3b <+11>: movsd 0x5aacd(%rip), %xmm0 ; typeinfo name for Sphere + 12, xmm0 = mem[0],zero
0x10002fc43 <+19>: movq 0x5c45e(%rip), %rax ; (void *)0x00007fff92ccdd40: __stack_chk_guard
0x10002fc4a <+26>: movq (%rax), %rax
0x10002fc4d <+29>: movq %rax, -0x8(%rbp)
0x10002fc51 <+33>: movl $0x0, -0x291c(%rbp)
0x10002fc5b <+43>: movl %edi, -0x2920(%rbp)
0x10002fc61 <+49>: movq %rsi, -0x2928(%rbp)
0x10002fc68 <+56>: movb $0x1, -0x2929(%rbp)
0x10002fc6f <+63>: movl $0x3c, -0x2930(%rbp)
0x10002fc79 <+73>: cvtsi2sdl -0x2930(%rbp), %xmm1
0x10002fc81 <+81>: divsd %xmm1, %xmm0
0x10002fc85 <+85>: movsd %xmm0, -0x2938(%rbp)
0x10002fc8d <+93>: leaq -0x2940(%rbp), %rdi
-> 0x10002fc94 <+100>: callq 0x100038310 ; std::__1::chrono::time_point<std::__1::chrono::steady_clock, std::__1::chrono::duration<long long, std::__1::ratio<1l, 1000000000l> > >::time_point at chrono:1338
0x10002fc99 <+105>: leaq -0x2948(%rbp), %rdi
0x10002fca0 <+112>: callq 0x100038310 ; std::__1::chrono::time_point<std::__1::chrono::steady_clock, std::__1::chrono::duration<long long, std::__1::ratio<1l, 1000000000l> > >::time_point at chrono:1338
0x10002fca5 <+117>: movq 0x5c384(%rip), %rdi ; (void *)0x00007fff92814760: std::__1::cout
0x10002fcac <+124>: leaq 0x44007(%rip), %rsi ; "Main Thread ID: "
0x10002fcb3 <+131>: callq 0x10007132c ; symbol stub for: std::__1::basic_ostream<char, std::__1::char_traits<char> >& std::__1::operator<<<std::__1::char_traits<char> >(std::__1::basic_ostream<char, std::__1::char_traits<char> >&, char const*)
It seems as the code just pics the first command it gets to crash.

Answer 1

The crash is due to a stack overflow that occurs when the gets too large. The simplest solution is to use a std::unique_ptr, because this array will be allocated on the heap.