Nikolai Pretzell wrote:
Stephan Bergmann wrote:
Nikolai Pretzell wrote:
[...]
Conclusion and suggestion
=========================
1)
The warnings don't give any additional information. Any attempt to
copy such a class results in a compiler error anyway.
The information is given when defining the class, not when using it,
so the warning can be of value when creating an API
Yes, in case of an API that does not plan and test, if the offered
classes are copyable, that warning would give a hint.
2)
The warnings annihilate the boost noncopyable-idiom
In how far do the warnings annihilate boost::noncopyable? The only
issue I see is that the workaround (declaring private, undefined
functions) and boost::noncopyable express the same thing, so using
them both is redundant.
The problem is, existing code using the boost idiom (or in another way
relying on the feature that one uncopyable member prevents the whole
class from being copyable) is forced to use the latter way, so making
the derivation from boost::noncopyable obsolete.
That means: We would break existing code that uses a quasi-standard
idiom or - as Thorsten described it - just a known language feature.
I think: If one writes APIs and thinks he needs these warnings, switch
them on explicitely in the affected modules.
But I do not want to break existing code that uses a widely accepted
idiom with a warning that for non-API code provides no information.
Nikolai
As there are two proponents, nobody who vetos loudly, and just one who
is feeling a little indifferent about it: I will add
-wd4511 -wd4512 -wd4625 -wd4626
to CFLAGSWALLCXX on wntmsci10.
-Stephan
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
