home
> blog
>> 796

translate to German (by SYSTRAN)

Weblog RDF feed, Atom feed

Mon Apr 22 22:10:43 2013 GMT: Die Zauberflöte @ Royal Opera House

Sat Apr 20 19:10:54 2013 GMT: C++14 Features

Sat Apr 20 14:40:44 2013 GMT: No Digit Separators in C++14

Yesterday it looked like digit separators would just about make it into C++14, but at the plenary session today there was no longer consensus for the proposal (Doug Gregor thankfully came up with some convincing examples of breakages compared to C++11) - WG21 voted 2 (favour) - 3 (opposed) - 3 (abstain).

Fri Apr 19 20:19:08 2013 GMT: C++14 Taking Shape

By Saturday evening it should be reasonable clear what features will be included in C++14 as the Committee Draft of C++14 will be produced after this week's Bristol meeting. "Digit separators" appears to cause quite some controversy with PL22.16 being opposed, but WG21 as a whole being in favour of the current proposal. The other controversial proposal (as it will break existing C++11 code) that appears to just have enough support is to make constexpr member functions no longer implicitly const.

Thu Apr 11 07:06:05 2013 GMT: auto - What type do you get?

If you have been to Roger's talk about auto and got confused by what type might actually get deduced, I have one more example to add to the confusion:

const int bar();
 
void foo()
{
  auto && var = bar();
  decltype(var)::error;
}

This was one of the examples he presented in his talk (and where the Visual Studio IDE couldn't agree with the Visual C++ compiler - Visual Studio thought it would be "int &&" and the compiler thought it would be "const int &&").

Now, you can also try this with a type other than int:

struct C
{ };
 
template const T bar();
 
void foo()
{
  auto && var1 = bar();
  decltype(var1)::error;
  
  auto && var2 = bar();
  decltype(var2)::error;
}

So what types do you get for var1 and var2?

Well, actually you should get "int &&" for var1 and "C const &&" for var2. Why do you get a "const" for one case and not for the other?

The reason here is because the C++ Standard says in 5/p6 "If a prvalue initially has the type "cv T," where T is a cv-unqualified non-class, non-array type, the type of the expression is adjusted to T prior to any further analysis."

Thu Apr 11 06:45:30 2013 GMT: DHCPv6 really working now on my Raspberry Pi

Actually, there was still a bug in dnsmasq 2.66rc1 which we managed to track down and Simon fixed it in 2.66rc5.

Sun Apr 07 20:38:11 2013 GMT: DHCPv6 on the Raspberry Pi

Just got DHCPv6 working now on my Raspberry Pi after upgrading to dnsmasq 2.66rc1 (raspian's dnsmasq 2.62 didn't pick up the statically configured IPv6 address for some reason).

Sat Apr 06 21:27:13 2013 GMT: Carmina Burana @ Royal Festival Hall

Fri Apr 05 23:04:07 2013 GMT: La Bayadère @ Royal Opera House

Mon Apr 01 18:16:04 2013 GMT: Weekend in Amsterdam

Sat Mar 30 19:06:27 2013 GMT: ISO C++ issue lists imported into roundup trackers

Sat Mar 09 10:17:43 2013 GMT: Written on Skin @ Royal Opera House