Re: [PATCH 0/1] mm: make read-only accessors take const pointerparameters

[Phil Carmody]

On 15/04/11 18:12 +0200, ext Michal Nazarewicz wrote:
> On Fri, 15 Apr 2011 17:59:16 +0200, Phil Carmody wrote:
>> I'm just glad this wasn't an insta-nack, as I am quite a fan of
>> consts, and hopefully something can be worked out.
>
> I feel you man.  Unfortunately, I think that const, since it's an
> after-thought, is not very usable in C.
>
> For instance, as you've pointed in your patch, the "_ro" suffix
> is sort of dumb, but without it compound_head would have to take
> const and return non-const (like strchr() does) which is kinda
> stupid as well.
>
> What's more, because of lack of encapsulation, âconst struct pageâ
> only means that the object is const but thighs it points to aren't.
> As such, const does not really play that well with structs anyway.

I'm very glad you've mentioned that point, I forgot to. I've taken
the view that in the absense of inside knowledge, const should be
inherited down all pointers. So not only will I not change you, but
I will not change anything you point to. No hidden side effects of
any kind. That reduces where it can be used, but is a much stronger
statement when it can be made.

> const is, in my opinion, one of those things C++ actually got
> right (or close to right).

I shouldn't be seen to agree with you on that, lest any fellow Nokians
notice that I've implied something positive about C++ after my rant on
our core chat channel a few days back. ;-)

Cheers,
Phil
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/