aaron.ballman added a comment. In D132997#3763367 <https://reviews.llvm.org/D132997#3763367>, @tbaeder wrote:
> In D132997#3763094 <https://reviews.llvm.org/D132997#3763094>, @shafik wrote: > >> I think like @aaron.ballman was saying in another PR you should aim for as >> complete set of tests as possible even if you have to comment one that can't >> work yet. For example cases that would involve `MemberExpr` or `UsingShadow` >> for example as well as the cases you mentioned not implemented in your >> description. > > I get it for the `MaterializeTemporaryExpr`s, but I don't understand why I > would add tests using `MemberExpr` now and not when I work on record types > (and references have already landed). The test case I was thinking of for that is: struct S { int i, j; }; constexpr int test() { S s{1, 2}; int &j = s.i; j += 10; return j; } static_assert(test() == 11, ""); where the reference is a `DeclRefExpr`. Even if structures don't work yet, capturing it as a test case now means we don't have to remember to think about "what if there's a reference to this member" when structures are implemented. ================ Comment at: clang/test/AST/Interp/references.cpp:13 + int a = 20; + int &b =a; + ---------------- CHANGES SINCE LAST ACTION https://reviews.llvm.org/D132997/new/ https://reviews.llvm.org/D132997 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
