As with many others commenting here, I've certainly had many suspicions I'd found a compiler (or equivalent) bug almost always proved false.
But in 30+ years of professional experience, I've also found two compiler-like bugs (I tend to use scripting / interpreted languages, so "compiler" isn't entirely accurate). One was in a commercial software package in which the documentation and implementation of a feature were reversed (what resolved as "true" should have been "false" and vice versa). That resulted in a code fix.
And another was a bug (specifics of which I've since forgotten) in GNU Awk, and not, I painstakingly verified, in my own code. That was also submitted and fixed.
But in 30+ years of professional experience, I've also found two compiler-like bugs (I tend to use scripting / interpreted languages, so "compiler" isn't entirely accurate). One was in a commercial software package in which the documentation and implementation of a feature were reversed (what resolved as "true" should have been "false" and vice versa). That resulted in a code fix.
And another was a bug (specifics of which I've since forgotten) in GNU Awk, and not, I painstakingly verified, in my own code. That was also submitted and fixed.
Every other time, though, my own damned fault ;-)