JonChesterfield marked an inline comment as done. JonChesterfield added inline comments.
================ Comment at: clang/test/CodeGenCXX/attr-no-zero-initializer.cpp:40 +// CHECK: @unt = global %struct.nontrivial undef +nontrivial unt [[clang::no_zero_initializer]]; ---------------- Quuxplusone wrote: > Can you explain a bit about how this interacts with C++ constructors? Will > this object not have its constructor called at startup; or is it that the > constructor will be called but the memory simply won't have been zeroed > //before// calling the constructor? > > For [P1144 relocation](https://wg21.link/p1144) (D50114, D61761, etc), Anton > Zhilin and I have been discussing the possibility of a similar-sounding > attribute that would skip the constructor of a local variable altogether, > allowing someone to write e.g. > ``` > T relocate(T *source) { > [[unconstructed]] T result; > memcpy(result, source, sizeof(T)); > return result; > } > ``` > If your attribute does exactly that, then I'm interested. If your attribute > doesn't do that, but is occupying real estate that //implies// that it does, > then I'm also interested. This change is orthogonal I think. No change to object lifetime. Without the attribute, a global is zero initialized and then the constructor is called before main. With it, the global is undef before the constructor call. A different attribute that suppresses the con(de)structor call entirely is also of interest to me. A freestanding implementation is likely to accept global constructors without warning, and emit code for them, but not actually run them at startup. That's not a great UX. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D74361/new/ https://reviews.llvm.org/D74361 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
