• Steve Teale (27/27) Jan 27 2014 I'm sure this was discussed at great length sometime, but
• bearophile (8/12) Jan 27 2014 A better compiler design is to have warning activated on default
• Dicebot (5/8) Jan 27 2014 Based on my C++ memories it is quite common issue when working in
• Steve Teale (5/9) Jan 27 2014 That is of course true. The worst case then being that other

"Steve Teale" <steve.teale britseyeview.com> writes:

I'm sure this was discussed at great length sometime, but 
yesterday (and the day before) I had one of those days where I 
turned on warnings , and told the compiler not to ignore 
deprecated things.

Most of the warnings were probably due to keyboard repeats where 
I had held a key down fractionally too long while typing the 
semicolon to end whatever. The other common case was fall-through 
in case statements - in that case, very insidious, and the 
warnings very helpful

All of the deprecated things in in my cleanup were those many 
cases where I had provided alternative definitions for virtual 
functions in derived classes without using override.

Now I can vaguely see the reasoning for insistence on the 
override keyword. It's kind of a consequence of Walter's decision 
that all member functions are virtual unless ...

But as I plodded through my code base, I came across not a single 
situation where the declaration of a virtual function in some 
derived class was an error.

I found a few where the declarations were redundant, in that they 
were the same in effect as the function defined in the parent 
class, but I was left feeling cheated. Being picky about the code 
had not given me any tangible benefit.

Would it be reasonable to ask for the compiler to generate a 
warning if it found that the definition of a virtual function was 
the same - disregarding white space - as that in the parent 
class. That would make me feel that there was some small benefit 
for the insistence on 'override'.

"bearophile" <bearophileHUGS lycos.com> writes:

Steve Teale:

 I'm sure this was discussed at great length sometime, but 
 yesterday (and the day before) I had one of those days where I 
 turned on warnings , and told the compiler not to ignore 
 deprecated things.

A better compiler design is to have warning activated on default 
and to be disabled on request with a compiler switch.

This is the enhancement request where I asked the obligatory 
usage of override:
http://d.puremagic.com/issues/show_bug.cgi?id=3836

Bye,
bearophile

"Dicebot" <public dicebot.lv> writes:

On Monday, 27 January 2014 at 16:06:46 UTC, Steve Teale wrote:
 But as I plodded through my code base, I came across not a 
 single situation where the declaration of a virtual function in 
 some derived class was an error.

Based on my C++ memories it is quite common issue when working in 
a teams because "Big Picture" is lacking. When you are main sole 
contributor, keeping overall class layout in mind does not seem 
to be any demanding.

"Steve Teale" <steve.teale britseyeview.com> writes:

 Based on my C++ memories it is quite common issue when working 
 in a teams because "Big Picture" is lacking. When you are main 
 sole contributor, keeping overall class layout in mind does not 
 seem to be any demanding.

That is of course true. The worst case then being that other 
users add functions with colliding signatures without being aware 
of the same method in some remote base class.

So yes, override has to be there, but I want my pound of flesh 
also ;=)