[cfe-dev] [RFC] Introduce overflow builtins

[Eli Friedman]

On Sat, Apr 21, 2012 at 8:21 AM, Tobias Grosser <tobias at grosser.es> wrote:
> On 04/04/2012 10:48 PM, Chris Lattner wrote:
>>
>>
>> On Apr 2, 2012, at 11:57 AM, Xi Wang wrote:
>>
>>> On Apr 2, 2012, at 1:04 PM, Dave Zarzycki wrote:
>>>
>>>> A colleague pointed out that i8 works. See the example at the end of
>>>> this email and please consider adding support for 8-bit integers for the
>>>> sake of completeness. Thanks! :-)
>>>
>>>
>>> Hah, thanks for letting me know this hidden feature.  Probably
>>> http://llvm.org/docs/LangRef.html should get updated, too.
>>>
>>> Pushed out.
>>>
>>> https://github.com/xiw/clang/compare/master...builtin-overflow
>>
>>
>> This looks generally reasonable to me, but could use review by someone
>> else.  Eli or John, could you take a look?
>
>
> I am very interested in this. The patch is very simple and looks at a first
> glance clean.
>
> Eli or John: I attached it to simplify review. Mind having a look?

Sorry about the long delay... I've been busy with other things, and
this slipped off my queue.

+  switch (Context.getTypeSizeInChars(ValTy).getQuantity()) {
+  case 1:
+  case 2:
+  case 4:
+  case 8:
+  case 16:
+    break;
+  default:
+    return Diag(DRE->getLocStart(), diag::err_overflow_builtin_pointer_size)
+      << Ptr->getType() << Ptr->getSourceRange();
+  }

I'm not sure what you're trying to catch with this check... it will
pass for every integer type we currently support.

Related: we probably want to reject __builtin_add_with_overflow(bool*,
bool, bool).

+    Intrinsic::ID ID = T->hasSignedIntegerRepresentation()? SID: UID;

Minor nit: whitespace.

Otherwise, this looks good.  Again, sorry about the delay.

-Eli