In C++0x is there something like static_assert which gives a warning instead of an error?

Question

I would like to do this for usages which may be inefficient but not necessarily incorrect.

Nor should there be in most cases: treat warnings as errors. — GManNickG, May 24 '11 at 08:58

Lightness Races in Orbit · Answer 1 · 2011-05-24T12:06:18.657

6

No.

An assertion failure indicates a problem preventing the program from being completed (be that execution [run-time assertions], or compilation [static assertions]).

In truth, an implementation is allowed to do anything as long as they emit a diagnostic (including continuing execution). But, in practice, mainstream toolchains will all behave pretty much the same: they will error out. You certainly can't hack them to something user-defined.

edited May 24 '11 at 12:06

answered May 24 '11 at 09:15

Lightness Races in Orbit

378,754
76
643
1,055

compile-time assertion failure prevents the code from *compiling*. running is out of question. – Nawaz May 24 '11 at 09:20
@Nawaz: I was using a more general meaning of "to run". Poor choice of word, perhaps. Allow me to pick a different one. – Lightness Races in Orbit May 24 '11 at 09:21
5

@Tomalak, @Nawaz - Although, if I may indulge in a bit of a language lawyering, the Standard never requires the compilation to halt if the compiler encounters ill-formed code (including a failed `static_assert`) -- only that a diagnostic message must be emitted. After this, the compiler is free to do whatever it wants, including finishing the compilation anyway. – JohannesD May 24 '11 at 09:51
1

@Tomalak : @JohannesD is correct. Check out [this](http://stackoverflow.com/questions/5177167/why-does-my-c-compiler-allow-recursive-calls-to-main/5177289#5177289) answer. A compiler is free to do whatever it wants to do. – Prasoon Saurav May 24 '11 at 11:48
@Prasoon: Oh, believe me, it's not that I don't trust him. – Lightness Races in Orbit May 24 '11 at 12:05

score 3 · Accepted Answer · answered May 24 '11 at 11:41

3

The attributes are introduced in C++0x for that purpose. See http://docwiki.embarcadero.com/RADStudio/en/C%2B%2B0x_attribute_deprecated for an example.

answered May 24 '11 at 11:41

vines

5,160
1
27
49

score 0 · Answer 3 · edited May 24 '11 at 11:35

0

Not as Standard, no. You can find #warning in many compilers, but that's really not the same in most situations.

edited May 24 '11 at 11:35

一二三

21,059
11
65
74

answered May 24 '11 at 08:58

Puppy

144,682
38
256
465

I thought `pragma warning` was to disable warnings? – Björn Pollex May 24 '11 at 09:00
1

And anyway, I've yet to come across a compiler that warns about "inefficiency". – May 24 '11 at 09:01
The problem is that `#warning` is selected by preprocessor conditions, which have extremely little ability to check other program content.... – Tony Delroy May 24 '11 at 10:11

In C++0x is there something like static_assert which gives a warning instead of an error?

3 Answers3