aganea added inline comments.
================
Comment at: clang/lib/Sema/SemaDeclCXX.cpp:11545
+ for (CXXMethodDecl *M : WorkList) {
+ DefineImplicitSpecialMember(*this, M, M->getLocation());
+ ActOnFinishInlineFunctionDef(M);
----------------
rnk wrote:
> hans wrote:
> > rnk wrote:
> > > Do we need to do this until fixpoint? Suppose a dllexported implicit
> > > special member triggers a template instantiation, and the template has a
> > > dllexported defaulted special member, something like:
> > > ```
> > > struct Bar { Bar(); };
> > > template <typename T> struct Foo { __declspec(dllexport) Foo() = default;
> > > Bar obj; };
> > > struct Baz {
> > > ... not sure how to trigger instantiation
> > > };
> > > ```
> > I think that should work, and that's why the function is written to be
> > re-entrant by having a local worklist. If it triggers a template
> > instantiation, ActOnFinishCXXNonNestedClass should get called for the newly
> > instantiated class. But I'm also not sure how to write a test case that
> > would trigger it.
> Right, I see. Nevermind then.
This works on latest MSVC:
```
$ cat test.cc
struct Bar { Bar(); };
template <typename T> struct Foo { __declspec(dllexport) Foo() = default; Bar
b; };
template class Foo<int>;
$ cl /c /Z7 test.cc <-- crashes if using clang-cl
$ lib test.obj /def
$ cat test2.cc
struct Bar { Bar() { } };
template <typename T> struct Foo { __declspec(dllimport) Foo() = default; Bar
b; };
int main() {
Foo<int> f;
return 0;
}
$ cl /c /Z7 test2.cc
$ link test.lib test2.cc /DEBUG
```
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D65511/new/
https://reviews.llvm.org/D65511
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
