OK, thanks.
Jason
Hi,
On 06/03/2014 05:41 PM, Jason Merrill wrote:
On 06/03/2014 11:24 AM, Paolo Carlini wrote:
+ if (NULLPTR_TYPE_P (from) && (flags & LOOKUP_ONLYCONVERTING))
+conv->bad_p = true;
Thanks. What kind of error message do we get with this change? Would
adding something to convert_li
On 06/03/2014 11:24 AM, Paolo Carlini wrote:
+ if (NULLPTR_TYPE_P (from) && (flags & LOOKUP_ONLYCONVERTING))
+ conv->bad_p = true;
Thanks. What kind of error message do we get with this change? Would
adding something to convert_like_real provide a more helpful diagnostic?
Hi,
On 06/03/2014 05:02 PM, Jason Merrill wrote:
Right. In the case of
+ TDerived()
+ : TBase(nullptr) { }
we have direct-initialization of TBase, but the parameter of the
TBase constructor is copy-initialized, so nullptr32.C is ill-formed;
please drop this hunk of the patch.
Oops, now
On 06/03/2014 10:30 AM, Paolo Carlini wrote:
implementing the resolution seems rather straightforward, just check
LOOKUP_ONLYCONVERTING in standard_conversion.
Yep. Though it would be better to return a bad_p conversion than none
at all.
However, while playing
with some additional tests ou
.. wondering if I should check DECL_CONSTRUCTOR_P (fn) too.
Paolo.
Hi,
implementing the resolution seems rather straightforward, just check
LOOKUP_ONLYCONVERTING in standard_conversion. However, while playing
with some additional tests outside bug & testsuite (similar to
nullptr32.C), I noticed a latent issue: in case of base initializers we
were setting any
