std::vector empties from one Line of Code to the next

Question

I'm currently programming something needing vectors. I have two nested loops;

for (size_t i = 0; i < m_neurons.size(); i++) {
        for (size_t j = 0; j < m_neurons[0].m_weights.size(); j++) {

        }
}

This section is the one we are interested in:

    d.out ("i == " + std::to_string (i));
    d.out ("m_neurons.size() == " + std::to_string (m_neurons.size()));
    d.out ("m_neurons[i].m_weights.size() == " + std::to_string (m_neurons.at(i).m_weights.size()));
    d.out ("m_neurons[i].m_weights.size() == " + std::to_string (m_neurons.at(i).m_weights.size()));
    for (size_t j = 0; j < m_neurons[0].m_weights.size(); j++) {

I print out some sizes and then, I want C++ to loop through the vector; easy enough, right?

Except this happens:

i == 0
m_neurons.size() == 1
m_neurons[i].m_weights.size() == 4
m_neurons[i].m_weights.size() == 4
terminate called after throwing an instance of 'std::out_of_range'
  what():  vector::_M_range_check: __n (which is 0) >= this->size() (which is 0)
Aborted (Speicherabzug geschrieben)
[maximilian@new-host library]$ 

1) Somehow the vector gets emptied 2) Somehow, the collection of the data is treated like I used .at (int index)

I have:

• tried alternating between vector.at() and vector[]
• deleted the executable and recompiled (multiple times)
• rebooted my pc [dunno, might help]
• investigated with GDB (the vector changed value from one step to the other)
• Used my mediocre googling skills to search something similar.
• Played a bit with the content of those vectors

I know that:

• The vector is NOT changed in between
• It is exactly THIS line of code that bugs out [because I have a printing s tatement exactly after]
• I recall it working a bit earlier in the day

Now my Questions are:

1. Is this a known issue/feature?
2. What caused it?
3. How can I fix it?

EDIT:

I made some code that boils down the problem. Seems to work fine. The thing is: NOTHING at all changes between my code and the one there. What could be the thing that triggers the bug? As said: in dbg, the size of the vector changes from one step to the other

#include <iostream> 
#include <vector>
#include "debugMachine.h"

class bar {
    public:
    bar (std::vector<double> i) {
        foobarmiz = i;
    }
    std::vector<double> foobarmiz;
};

class foo {
    public:
    std::vector<bar> foobar;
    void test (std::vector<double>& expected) {
        d.out ("HI");
        for (size_t i = 0; i < foobar.size(); i++) {
            d.out ("HI2");
            d.out ("i == " + std::to_string (i));
            d.out ("m_neurons.size() == " + std::to_string (foobar.size()));
            d.out ("m_neurons[i].m_weights.size() == " + std::to_string (foobar.at(i).foobarmiz.size()));
            d.out ("m_neurons[i].m_weights.size() == " + std::to_string (foobar.at(i).foobarmiz.size()));
            for (size_t j = 0; j < foobar[0].foobarmiz.size(); j++) {
                d.info ("Gonna change weights");
                foobar.at(i).foobarmiz[j] +=1; 
            }
            d.out ("out of backward");
        }
    }
} f;

int main () {
    d.write = true;
    std::vector<double> test {1.0f, 2.0f, 3.0f, 5.34f};
    f.foobar.push_back (bar (test));
    f.foobar.push_back (bar (test));
    f.foobar.push_back (bar (test));
    f.test (test);
}

Here is the entirity of the function that bugs out:

void CLayer::m_backward (std::vector<double>& expected) {
    d.out ("HI");
    for (size_t i = 0; i < m_neurons.size(); i++) {
        d.out ("HI2");
        d.out ("i == " + std::to_string (i));
        d.out ("m_neurons.size() == " + std::to_string (m_neurons.size()));
        d.out ("m_neurons[i].m_weights.size() == " + std::to_string (m_neurons.at(i).m_weights.size()));
        d.out ("m_neurons[i].m_weights.size() == " + std::to_string (m_neurons.at(i).m_weights.size()));
        for (size_t j = 0; j < m_neurons[0].m_weights.size(); j++) {
            d.info ("Gonna change weights");
            d.info ("m_weights before : " + std::to_string(m_neurons.at(i).m_weights.at(j)));
            m_neurons.at(i).m_weights[j] -= (0.003f * -(m_expected.at(i) - m_output.at(i)) * m_output.at(i) * (1 - m_output.at(i)) * m_input.at(j)); 
            d.info ("m_weights after : " + std::to_string(m_neurons[i].m_weights[j]));
        }
        d.out ("out of backward");
    }
}

Answer 1

This is the sample that gives the problem.

#include <string>
#include <iostream> 
#include <vector>

class bar {
public:
    bar(std::vector<double> i) {
        foobarmiz = i;
    }
    std::vector<double> foobarmiz;
};

class foo {
public:
    std::vector<bar> foobar;
    void test(std::vector<double>& expected) {
        std::cout << ("HI");
        for (size_t i = 0; i < foobar.size(); i++) {
            std::cout << ("HI2") << std::endl;
            std::cout << ("i == " + std::to_string(i)) << std::endl;
            std::cout << ("m_neurons.size() == " + std::to_string(foobar.size())) << std::endl;
            std::cout << ("m_neurons[i].m_weights.size() == " + std::to_string(foobar.at(i).foobarmiz.size())) << std::endl;
            std::cout << ("m_neurons[i].m_weights.size() == " + std::to_string(foobar.at(i).foobarmiz.size())) << std::endl;
            for (size_t j = 0; j < foobar[0].foobarmiz.size(); j++) { // because of this line
                std::cout << ("Gonna change weights") << std::endl;
                foobar.at(i).foobarmiz[j] += 1;
            }
            std::cout << ("out of backward") << std::endl;
        }
    }
} f;

int main()
{
    std::vector<double> test{ 1.0f, 2.0f, 3.0f, 5.34f };
    f.foobar.push_back(bar(test));
    std::vector<double> test2{ 1.0f, 23.3f, 2.1f };
    f.foobar.push_back(bar(test2));
    std::vector<double> test3{ 1.0f, 23.3f };
    f.foobar.push_back(bar(test3));
    f.test(test);
}