llvm dev - Feb 2015 - [LLVMdev] Do we still need LLVM_HAS_INITIALIZER_LISTS?

On Mon, Feb 16, 2015 at 10:30 PM, David Majnemer
<david.majnemer at gmail.com> wrote:
>
>
> On Mon, Feb 16, 2015 at 3:56 PM, Justin Bogner <mail at
justinbogner.com>
> wrote:
>>
>> Currently, LLVM_HAS_INITIALIZER_LISTS is set based on
>>
>>   __has_feature(cxx_generalized_initializers) &&
!defined(_MSC_VER)
>>
>> Based on the docs, we should be able to drop the _MSC_VER check now
that
>> we've moved to VS2013. Can somebody confirm that?
>>
>
> MSDN claims that VS2013 supports braced initializer list syntax:
> https://msdn.microsoft.com/en-us/library/dn387583.aspx
>
>>
>> The other problem is that GCC doesn't recognize __has_feature, so
this
>> is essentially only enabled for clang. If we're keeping this check
it
>> should really check the GCC version as well - they've had these for
>> quite a while now.
I think we're safe to use it. I've made an experimental commit in
r229502 to see if any of the bots complain, and will revert if they
do. Otherwise, I think we're good to go.

~Aaron

On 2/17/2015 4:39 PM, Aaron Ballman wrote:
> I think we're safe to use it. I've made an experimental commit in
> r229502 to see if any of the bots complain, and will revert if they
> do. Otherwise, I think we're good to go.
>
One specific thing doesn't work: in-class initializers with brace syntax:

class Foo
{
   Bar bar{1, 2, 3};
};

They disabled this due to a silent wrong-code bug.

Sebastian

On Wed, Feb 18, 2015 at 7:23 AM, Sebastian Redl
<sebastian.redl at getdesigned.at> wrote:
>
> On 2/17/2015 4:39 PM, Aaron Ballman wrote:
>>
>> I think we're safe to use it. I've made an experimental commit
in
>> r229502 to see if any of the bots complain, and will revert if they
>> do. Otherwise, I think we're good to go.
>>
> One specific thing doesn't work: in-class initializers with brace
syntax:
>
> class Foo
> {
>   Bar bar{1, 2, 3};
> };
>
> They disabled this due to a silent wrong-code bug.
That's good to know, but also a coding style that we don't use (copy
initialization using = is preferred to direct initialization using {},
from what I learned recently). We should document both things in our
guidelines, though.

Thanks!

~Aaron