Since I started competing in competitive programming, I discover more and more ways to get RTE or WA verdicts because of some stupid mistakes I make during the coding phase, I will try to mention some of them here for my future self, and you can also share your experience in the matter.

Initially, I will list them without any specific order, and I may write a proper blog if I find myself with good material, so feel free to comment your thoughts.

    - wrong fixed arrays sizes // such as 3*10^5 and 10^6

    - #define int int64_t // work smart not hard

    - #define endl '\n' // one less thing to take care of

    - 1e9+9 // or any other stupid value

    - INF vs INFLL // always use INFLL with its value equal to 0x3f3f3f3f3f3f3f3f ("0x" + "3f"*(8)) // this can be used for int32 also

    - defining void function as a non-void one without returning any thing. // compiler fault and it mostly will works just fine in your machine, not like the judges one.

    - freopen the wrong file name // sad memories lives here

    - not printing the size of the answer when the problem asked for it 

    - scanf("%d", (long long)x)

    - for(int i = 2; i*i <= x; i++) if(!not_prime[i]) // i*i may overflow 

    - while(p%x==0) p/=x; // p may be equal to zero.

    - n*(n+1)/2 

    - lcm = x*y/gcd(x, y) // where (x=y=0)

    - gcd(x, 0) = x // special case 

    - (x-y)%mod // just used good defined adding and multiplying functions

    - (x+y-1)/y where y < 0 // there is no need to use the trick in negative numbers

    - x & (1<<p) // use ((x>>p)&1)

    - for(int i = empty_vec.size()-1; i >= 0; i--) // unsigned integer

    - for(int i = 0; i < log_n; i++) if(par[u][i] != par[v][i]) // (LCA) // start from the MSB to LSB, not the other way.

    - for(int i = 0; i < strlen(s); i++)

    - --upper_bound(vec.begin(), vec.end(), (pair<int, int>) {first, -1}) // use lower_bound({first+1, -1}) or inf instead of -1

    - lower_bound(set.begin(), set.end(), element) // you silly 

    - *empty_set.begin(), *--empty_set.end() // you're not trustworthy

    - x = min(x, map[x]) // map[new element] equal to zero.

    - multiset.erase(x)

    - not using size variable in centroid decomposition, hld or small-to-large // very dangerous one

    - DFSing on other than the 0-th node when filling binary-lifting values (LCA). // par[u][inf]=0

    - using a wrongly sorted queue in Dijkstra algo. // multiply by -1 instead of using a greater operator

    - not re initiating global variables in multi testcases problem. 

    - using memset or fill in a multi test cases problem // be smart and use integer-visited-arrays with visited-id

    - breaking from the user input loop in a multi test cases problem

    - segment tree base cases conditions // (r < q_l || q_r < l)  // don't trust me with this :)

    - using 1-indexed bit or sparse table. // never use 1-indexed stuff

    - [l, r) // who define things this way??

    - ~node() // not writing a destruction function for a pointer based segment tree or trie in a multi testcase problem.

    - new node() // MLE

    - if(x) where x < 0

    - if(x==true && y==true) // put parentheses (acpc kickoff 2020)

    - if(x&(1<<p)==0 && y==true) // & is evil 

    - b = y&(1<<p); x = child[y][b] // b must be a boolean not 2^p

    - to_string // write it yourself

    - std rand function // use mt19937

    - const bool debug = true // check any debug-related stuff like watchers 

    - Yes vs YES

    - alic and bobe

    - n==1 or even n==0

    - use array<> instead of vector<> when size is fixed

    - return answer // return mem[x] = answer

    - return x <= y // in user defined compair functions

    - switching between n and m // 15m of awkwardness in a session // spoj.com/problems/LABYR1

    - vector.clear() // use v = vector<int>(), because .clear() or .resize(0) doesn't deallocate memory