[PATCH] D147591: [clang][Interp] Handle CXXTemporaryObjectExprs

[Aaron Ballman via Phabricator via cfe-commits]

aaron.ballman accepted this revision.
aaron.ballman added a comment.
This revision is now accepted and ready to land.

LG pending test coverage being ready to land



================
Comment at: clang/test/AST/Interp/records.cpp:317-318
     {
-      auto T = Test(Arr, Pos);
+      Test(Arr, Pos);
       // End of scope, should destroy Test.
     }
----------------
tbaeder wrote:
> aaron.ballman wrote:
> > tbaeder wrote:
> > > aaron.ballman wrote:
> > > > tbaeder wrote:
> > > > > aaron.ballman wrote:
> > > > > > Nit: nothing actually tests that this object is destroyed correctly. Here's an interesting test to consider:
> > > > > > ```
> > > > > > struct S {
> > > > > >   constexpr S() {}
> > > > > >   constexpr ~S() noexcept(false) { throw 12; }
> > > > > > };
> > > > > > 
> > > > > > constexpr int f() {
> > > > > >   S{};
> > > > > >   return 12;
> > > > > > }
> > > > > > 
> > > > > > static_assert(f() == 12);
> > > > > > ```
> > > > > > That should fail because `~S()` would hit the `throw` expression and thus is not valid. Note, you'll need to add `-Wno-invalid-constexpr` to your test to avoid the warning-defaults-to-error about the destructor never producing a constant expression.
> > > > > There are multiple reasons why that sample is not rejected right now, one I can easily fix in a follow-up patch, the other one would actually require us to recognize the `throw` and reject it with a proper diagnostic.
> > > > We should definitely fix the `throw` at some point, but any of the dynamically reachable problematic constructs would work (`dynamic_cast` whose type would throw, invocation of the `va_arg` macro, `reinterpret_cast`, etc)
> > > Yes, I think we need a new opcode for that so we only emit the diagnostic when such a construct is actually executed.
> > Oh yeah, you'll definitely need that, a whole pile of the constexpr rules are based around code reachability.
> > 
> > Are you saying you've got no way to test this until you implement that opcode?
> With https://reviews.llvm.org/D150040 applied, it gets properly rejected, just the diagnostics are off. I can add the test and reorder the commits.
SGTM, thanks!


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D147591/new/

https://reviews.llvm.org/D147591