Here is the discussion of the original Trusting Trust attack: https://news.ycombinator.com/item?id=13569275

Thanks I was referring to precedence climbing parser part of GP reply.

Thompson publicized an attack by Paul Karger during MULTICS. One of many. Defeating their totality took what's called high-assurance security. For compilers, you have to verify source despite potentially-malicious developers, verify its binary translation, and its distribution. Maybe the tools used to do that as well. I wrote more about what that takes here:

https://lobste.rs/s/nognrl/creating_language_using_only_asse...