When researching an answer to a question (based on this answer) I tried to do the following:
在研究问题的答案时(基于这个答案),我尝试做以下事情:
template <class T>
class friendly {
friend class T;
};
friendly<string> howdy;
This fails to compile with the following error:
无法编译时出现以下错误:
error: template parameter "T" may not be used in an elaborated type specifier friend class T;
错误:模板参数“T”不能用在精心设计的类型说明符朋友类T中;
From what I can understand from my good friend Google this is so that I won't accidentally try to instantiate friendly<int>
but why should this be an error when compiling the template? Should't it be an error when instantiating the template with an invalid type (such as had I written int f() { return T::foo(); }
)
从我的好朋友Google可以理解的是,我不会不小心尝试实例化友好的
2 个解决方案
#1
Section 7.1.5.3 of the standard explicitly describes this as an example of an illformed elaborated type specifier.
该标准的7.1.5.3节明确地将此描述为一个错误的详细类型说明符的示例。
A discussion about the subject can be found here.
关于这个问题的讨论可以在这里找到。
#2
A bit more googleling brought up Extended friend
Declarations (PDF) for C++0x.
更多googleling带来了C ++ 0x的扩展好友声明(PDF)。
This document contains the following:
本文档包含以下内容:
template <typename T> class R {
friend T;
};
R<C> rc; // class C is a friend of R<C>
R<int> ri; // OK: “friend int;” is ignored
Which goes even further than what I thought (ignoring illegal friend
decelerations rather than failing during instantiation). So I guess the answer is that there isn't a good reason and it's being rectified.
这比我的想法更进一步(忽略非法的朋友减速而不是在实例化期间失败)。所以我想答案是没有充分的理由而且它正在被纠正。
#1
Section 7.1.5.3 of the standard explicitly describes this as an example of an illformed elaborated type specifier.
该标准的7.1.5.3节明确地将此描述为一个错误的详细类型说明符的示例。
A discussion about the subject can be found here.
关于这个问题的讨论可以在这里找到。
#2
A bit more googleling brought up Extended friend
Declarations (PDF) for C++0x.
更多googleling带来了C ++ 0x的扩展好友声明(PDF)。
This document contains the following:
本文档包含以下内容:
template <typename T> class R {
friend T;
};
R<C> rc; // class C is a friend of R<C>
R<int> ri; // OK: “friend int;” is ignored
Which goes even further than what I thought (ignoring illegal friend
decelerations rather than failing during instantiation). So I guess the answer is that there isn't a good reason and it's being rectified.
这比我的想法更进一步(忽略非法的朋友减速而不是在实例化期间失败)。所以我想答案是没有充分的理由而且它正在被纠正。