Sanitizers and Standard Won’t Save Us

Some consider C++ to be a great language due to its abundance of literature and courses, as well as sanitizers for detecting undefined behavior, powerful IDEs and static analysis tools. But how robust are these? What happens when a programmers rely on them but cannot recall the Standard from their memory? Is it helpful to require full correctness with five compilers in 15 modes across 3 operating systems? What can we uncover by using C++ only for training exercises? Are there any holes in the Standard, peculiarities of compilers, bugs in tooling?

This talk covers some (non-)foolproof mechanics in operating systems, version control systems, GitHub Actions-based infrastructure, as well as incompatibilities and Standard violations in compilers and tooling. All bugs were discovered by several generations of freshmen students.