labath added a comment. In D81471#3637057 <https://reviews.llvm.org/D81471#3637057>, @werat wrote:
> In D81471#3632297 <https://reviews.llvm.org/D81471#3632297>, @labath wrote: > >> I guess the condition we really want to express here is "does this >> expression refer to a constexpr variable (ideally one without a location)"? >> And the problem is that clang does not give us the means to detect that? >> >> Is that really the case? Would it maybe be possible to use some c++ magic to >> get clang to do that for us. Write something like `if constexpr >> (__builtin_constant_p(user_expression)) do_something_rvalue_like(); else >> assume_regular_lvalue();` ? > > I think you're right here, but I don't know a better way to express this. My > clangfu is not good enough for this yet :) > > As I understand it, we can't express it purely in the expression (via > `__builtin_constant_p` or something), because we need to know the answer > before executing the expression (it changes the way we do > materialization/dematerialization). Well, yes, either that, or change how the dematerialization works -- prepare for both options and then choose the one that was actually used. > Could this be something to improve in the future? Well.. maybe... I don't know. I think this is definitely useful, and I don't see much of a downside. The main difference I see is that you won't be able to "modify a constant" via expressions like `p const_int`, `p (int&)$0 = 47`, but: a) you will still be able to do that within a single expression; and b) such an operation is unlikely to produce predictable results in anything but pure C (which doesn't really have constants). So, if noone objects to this, then I think it's fine. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D81471/new/ https://reviews.llvm.org/D81471 _______________________________________________ lldb-commits mailing list lldb-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
