RFR: 8300080: offset_of for GCC/Clang exhibits undefined behavior and is not always a compile-time constant

[Johan Sjölen]

On Tue, 2 Sep 2025 23:25:43 GMT, Kim Barrett <kbarrett at openjdk.org> wrote:

> Please review this change to offset_of to just use offsetof as it's
> implementation. The rationale for this is
> 
> (1) Despite being UB for some cases that we use (through C++14), none of the
> compilers we use actually treat it as such, other than some compilers issuing
> warnings. (Surprisingly, the UB doesn't appear to disable constexpr evaluation?)
> 
> (2) C++17 changed it to be conditionally defined in those formerly UB cases,
> and we'll be switching over to C++17 soon (JDK-8314488).
> 
> Until we make the switch to C++17 we need to suppress warnings from some
> compilers for such uses. We can remove that warning suppression later.
> 
> See the JBS issue for more discussion.
> 
> Testing: mach5 tier1

Seems good to me. As Stefan hints at, maybe it's worth waiting until C++17 integration and then simplify the PR to only change the macro?

-------------

Marked as reviewed by jsjolen (Reviewer).

PR Review: https://git.openjdk.org/jdk/pull/27061#pullrequestreview-3180362344