[libcxx-commits] [libcxx] [libcxx] fix mi_mode_test failure with libc++-21 (PR #153969)

[Sv. Lockal via libcxx-commits]

AngryLoki wrote:

I guess I understand the idea. 
Test subapplication (built with some stdlib) has `std::unordered_{map,set}` hashmap.
GDB provides its dump (prints buckets in order they are filled). Then test main application (built with **the same** stdlib) also creates `std::unordered_{map,set}`, iterates through the buckets. As long as hash table does not use hash randomization (as in Python, which has hash randomization), the order within **the same** stdlib should be the same.
I'll update this PR tomorrow.

https://github.com/llvm/llvm-project/pull/153969