Repository navigation
Inconsistent class-key in class template declarations #534
Description
Declarations such as e.g.
// 20.3.4, tuple-like access to pair:
template <class T> class tuple_size;
template <size_t I, class T> class tuple_element;
template <class T1, class T2> struct tuple_size<pair<T1, T2> >;
template <class T1, class T2> struct tuple_element<0, pair<T1, T2> >;
template <class T1, class T2> struct tuple_element<1, pair<T1, T2> >;Use struct and class interchangeably. While this is not a defective thing per se (see §7.1.6.3/3), compilers do warn about inconsistent usage of both and it should be tidied up.
Activity
jwakely commentedon Aug 27, 2015
Compilers don't compile declarations in the standard.
Arcoth commentedon Aug 27, 2015
My point is that the aforementioned usage can be irritating. Also, yes they do, as the declarations are put into headers by compiler-devs that themselves are compiled when including them - unless I'm mistaken? (Of course, standard headers don't produce warnings, but a specialization of tuple_size necessarily does, which is annoying!)
jwakely commentedon Aug 27, 2015
Compilers do not compile the standard, they compile headers that come with their implementation. The implementation headers can use whatever class-keys they want to, they don't have to match the standard, and changing the standard doesn't mean implementations will change.
I see nothing worth fixing here.
oblitum commentedon Aug 27, 2015
Nor do I, I use
classandstructinterchangeably (I don't likeclasswithstruct-like layout for example, with noprivatemembers, etc). If any compiler I use were to warn about that I would just use a compiler flag to silence such warning.sigfpe commentedon Aug 27, 2015
Users who tolerate those bogus warnings should find something more productive to do.
-------- Original message --------
From: Francisco Lopes notifications@github.com
Date: 08/27/2015 9:54 AM (GMT-08:00)
To: cplusplus/draft draft@noreply.github.com
Subject: Re: [draft] Inconsistent class-key in class template declarations
(#534)
Nor do I, I use class and struct interchangeably (I don't like class with struct-like layout for example, no private members, etc). If any compiler where to warn about that I would just use a flag to silent such warning.
—
Reply to this email directly or view it on GitHub.
zygoloid commentedon Aug 27, 2015
Closing; while consistency in presentation is generally important, in this case I don't think the comprehensibility or style of the standard is affected.
oblitum commentedon Aug 27, 2015
I've captured the original intent of the issue re-reading it now... Well in that case it seems understandable, but it's bogus nonetheless. Personally I use
classandstructinterchangeably with a purpose as I said. In this case deviation from the original template in the specialization doesn't look purposeful but there's no rule one can really form about that. Such warnings woudn't apply for me for example, since any specialization may happen to be moreclass-like orstruct-like, and I use the proper keyword in each case, regardless of the keyword used for the original template.Arcoth commentedon Aug 28, 2015
@zygoloid It might be sensible to uncouple
-Wmismatched-tagsfrom-Wallthen, for Clang. (You can do that, right?)Edit: Hmm, a bit of research reveals that this is partially relevant for linking, as the VS ABI appears to link
structandclassdifferently. Still seems that having to activate it explicitly is more sensible, though.jwakely commentedon Aug 28, 2015
@Arcoth, I still don't understand why you think the use of
structorclasshere in the standard has anything to do with warnings from your compiler. How we present this information in the standard does not dictate how vendors implement those specializations. If you get warnings complain to your std::lib vendor, but this is not the right venue.Arcoth commentedon Aug 28, 2015
I don't (anymore).
I was talking about Clang's nonsensical warning that annoyed me, which was the original incentive for opening this issue. I asked a Clang dev above whether removing it from
-Wallis sensible - but apparently OT isn't welcome here. I'll ask somewhere else then.sigfpe commentedon Aug 28, 2015
You do realize this is not the place to ask your vendor to do something very specific to your compiler.
-------- Original message --------
From: Arcoth notifications@github.com
Date: 08/28/2015 2:26 AM (GMT-08:00)
To: cplusplus/draft draft@noreply.github.com
Cc: sigfpe sig.fpe@axiomatics.org
Subject: Re: [draft] Inconsistent class-key in class template declarations
(#534)
I still don't understand why you think the use of struct or class here in the standard has anything to do with warnings from your compiler.
I don't.
If you get warnings complain to your std::lib vendor, but this is not the right venue.
I was talking about Clang's nonsensical warning that annoyed me, which was the original incentive for opening this issue. I asked a Clang dev above - apparently you're not welcoming OT. I'll ask somewhere else then.
—
Reply to this email directly or view it on GitHub.