wt., 2 maj 2023 o 17:09 Arthur O'Dwyer via Std-Proposals
<std-proposals_at_[hidden]> napisał(a):
>
> This is Clang's `-Wsuggest-override`. I always turn it on and keep my codebase clean of its warnings. No change to the Standard required.
>

This and possible add `[[my::override_strict]]` and use Clang-Tidy to
enforce all special rules.


> On Tue, May 2, 2023 at 10:21 AM Frederick Virchanza Gotham via Std-Proposals <std-proposals_at_[hidden]> wrote:
>>
>> Right now in my day job I'm trying to fix a bug that was introduced by
>> my predecessor. The code is C++ for a microcontroller. There is an
>> abstract base class, 'Motor'. Derived from that is 'FPGAMotor'.
>> Derived from that is 'StageMotor'.
>>
>> Now I think over the years we're all guilty of taking a short cut
>> rather than re-writing a class hierarchy properly. It turns out that a
>> lot of the FPGA stuff is actually in the base class, 'Motor', even
>> though a totally unrelated kind of motor could potentially inherit
>> from 'Motor'. I'm guessing that my predecessor just saw a quick fix
>> rather then spending hours re-arranging that class hiearchy.
>>
>> Now that I need to debug the code, I'm refactoring it and re-writing
>> the hierarchy properly. Every time I see a method defined inside a
>> class, there are a few possibilities:
>>
>> (1) The method is not inherited and is intended to be virtual
>> (2) The method is not inherited and is not intended to be virtual
>> (3) The method is inherited, and so it should be virtual in the base class(es)
>> (4) The method is inherited, and is expected to have no prior
>> definition (i.e. it's expected to be pure virtual in the base
>> class(es))
>> (5) The method is inherited, but it's not virtual in the base class
>> and isn't intended to be virtual in the derived class either.
>>
>> Scenario No. 5 is usually a red flag, and I wish the compiler would disallow it.
>> Scenario No. 1 and No. 2 are both valid and common.
>> Scenario No. 3 is the most common (this is what the 'override' keyword is for)
>> Scenario No. 4 is less common than No. 3 but is still valid
>>
>> Without breaking existing code, I'd like if a class could have a
>> 'strict' mode, something like:
>>
>> class MyClass strict {
>>
>> };
>>
>> When a class is written in 'strict' mode, each of the scenarios
>> described above must have the following or else you get a compiler
>> error.
>> Scenario 1) must be prefixed with 'new virtual'
>> Scenario 2) must be prefixed with 'new !virtual'
>> Scenario 3) must be prefixed with 'virtual', and also must be
>> postfixed with 'override'
>> Scenario 4) must be prefixed with 'virtual', and also must be
>> postfixed with 'override(0)'
>> Scenario 5) must be prefixed with '!virtual" and must be postfixed
>> with 'override(!virtual)'
>>
>> If the compiler would enforce these rules, it would be easier to clean
>> up old code.
>> --
>> Std-Proposals mailing list
>> Std-Proposals_at_[hidden]
>> https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>
> --
> Std-Proposals mailing list
> Std-Proposals_at_[hidden]
> https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals