you're using c++14 while they're using c++ 20

maybe its because the language? (one is C++ 14 and other is C++ 20)

gcc 6-32 uses uint32_t as size_t, but gcc 13-64 uses uint64_t.

The string::npos is defined as size_type npos = -1. Thus in the wrong submission, it is interpreted as $$$2^{32}-1$$$, and then cast into long long so either ia or ib will get $$$2^{32}-1$$$ instead of $$$-1$$$.

In the accepted submission, string::npos is interpreted as $$$2^{64}-1$$$, so ia or ib can get the expected $$$-1$$$.

To avoid this error, you can declare auto ia = s.find(..), and use string::npos instead of -1 explicitly.