One of my friends noticed a strange behaviour of .size() method.

Link to WA submission : https://codeforces.me/contest/1307/submission/88824759
Link to AC submission : https://codeforces.me/contest/1307/submission/88824749

The bug is in the line
LL x=a[i].size()*(a[i].size()-1)/2;

In C++17 , I guess there is overflow in that expression.
But in C++17(64) , it works fine.

So does that mean the .size() operator can hold larger numbers in 64bit version?
And also what other methods show this kind of varying behaviour when the compiler changes from 32bit to 64bit??
Can anyone explain this please??