a = a++ + ++a in C/C++ invokes undefined behavior twice, producing three compiler-valid answers: 11, 12, or 13.
Key Takeaways
Two UB sources: unspecified fetch order of a operands, and whether post-increment applies before or after assignment.
Cross-compiler table (gcc, clang, MSVC, Keil, SDCC, Java, C#) shows results splitting between 11-14 depending on toolchain and version.
Java and C# are fully defined: strict left-to-right evaluation means Java always returns 11, C# always returns 11.
C# spec §7.3 and Java JLS §15.7 explicitly guarantee evaluation order, making this a non-riddle in those languages.
Overloaded operators in C++ (g++, MSVC) can produce different results than built-in int for identical expressions.
Hacker News Comment Review
Consensus: the post understates UB severity – three possible answers implies bounded behavior, but true UB allows segfaults, nasal demons, or anything; no result is “correct.”
Commenters trace the root cause to C’s design goal of accommodating diverse hardware and dumb compilers, which made evaluation order intentionally unspecified to allow architecture-specific calling conventions.
The post resurfaces a real pedagogical harm: C/C++ textbooks (notably in India’s CS curriculum) treated these as exam questions with definite answers, misleading generations of students.
Notable Comments
@Boxxed: “if you invoke undefined behavior any result at all is possible” – the article’s three-answer framing is itself misleading.
@jcranmer: on gcc, argument evaluation order for f(a(), b()) is ABI-dependent, illustrating why C left this unspecified across architectures.
