================
Comment at: lib/Parse/ParseCXXInlineMethods.cpp:73
@@ -72,3 +72,3 @@
bool Delete = false;
SourceLocation KWLoc;
if (TryConsumeToken(tok::kw_delete, KWLoc)) {
----------------
rnk wrote:
> Any reason we can't save Tok.getEndLoc() here? Then we don't have to hardcode
> the lengths of delete and default.
Great suggestion, thanks. Done
================
Comment at: lib/Parse/ParseCXXInlineMethods.cpp:80
@@ -79,1 +79,3 @@
Delete = true;
+ if (auto *DeclAsFunction = dyn_cast<FunctionDecl>(FnD)) {
+ // The end of this decl has to point to the end of the 'delete' keyword
----------------
rnk wrote:
> This doesn't do the right thing for function templates, which the only case
> where this dyn_cast can fail. I think you can write a templated constructor
> and default it. =P A robust way to do this might be:
> auto *FD = dyn_cast<FunctionDecl>(FnD);
> FD = FD ? FD : cast<FunctionTemplateDecl>(FnD)->getTemplatedDecl();
> FD->setRangeEnd(KWEndLoc);
Reid, I may be missing something here, do you mean something like:
struct Foot {
template <typename T>
Foot &operator=(const T& Other) = default;
};
I don't think this is actually valid. The standard says in 8.4.2:
"A function that is explicitly defaulted shall ... have the same declared
function type (except for possibly differing ref-qualifiers and except that in
the case of a copy constructor or copy assignment operator, the parameter type
may be “reference to non-const T”, where T is the name of the member function’s
class) as if it had been implicitly declared,
and"
And indeed Clang complains:
error: only special member functions may be defaulted
Foot &operator=(const T& Other) = default;
http://reviews.llvm.org/D8465
EMAIL PREFERENCES
http://reviews.llvm.org/settings/panel/emailpreferences/
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
