[webkit-dev] Using C++ constant pointers (type_name * const) in WebKit

[Jarred Nicholls]

On Mon, Nov 28, 2011 at 5:19 PM, Ryosuke Niwa <rniwa at webkit.org> wrote:

> On Mon, Nov 28, 2011 at 1:38 PM, David Kilzer <ddkilzer at webkit.org> wrote:
>
>> In a discussion on Bug 71921<https://bugs.webkit.org/show_bug.cgi?id=71921>,
>> Antti, Darin Adler and I started a discussion about using C++ constant
>> pointers in WebKit.  Does the WebKit community have a consensus opinion on
>> the matter?
>>
>> * Pros
>>   - Documents use of variable.
>>   - Prevents misuse of variable in a later patch (by a different author)
>> through enforcement of pointer const-ness.
>>   - May help compiler optimize code.  (We weren't sure whether modern
>> compilers do this on their own or not.)
>>
>
> Don't think so. As far as I know, C++ compilers ignore "const" for the
> purpose of code optimization because we can always re-cast it to non-const
> variable. On the other hand, modern optimizing compilers might do SSA-based
> constant<http://en.wikipedia.org/wiki/Sparse_conditional_constant_propagation>
>  propagation <http://gcc.gnu.org/news/ssa-ccp.html> regardless of
> constness of a variable.
>
> * Cons
>>   - Darin Adler doesn't ever recall fixing a bug in WebKit where a
>> constant pointer would have helped.
>>   - Slightly more verbose syntax for constant pointers to a constant
>> string (const char * const pointer;) or even a constant pointer to a
>> mutable string (char * const pointer;).
>>
>
> I'd argue against using constant pointers simply because I don't think the
> existing WebKit code uses it, and I don't see much benefit in comparison to
> pointers or references to const objects.
>

I would agree with Ryosuke.  Other than the self-documenting benefits, it's
not as valuable as ref-to-const and pointer-to-const when one is passing
the variable around into different functions.

The bugs it would help prevent at compile-time are few and far between; too
much so to start enforcing the style right now.  But technically I don't
see the harm in letting them slip in where it makes sense.  From a
readability standpoint, they are very verbose and make it a little more
confusing to understand the intention during quick scans (ptr-to-const vs
const-ptr) - especially since ptr-to-const can be pre- or post- type.  *const
char** is the same as *char const** which is not the same as *char* const*.

I will say that in all the code I've written and read, const ptrs were
rarely used, if at all.


>
> - Ryosuke
>
>
> _______________________________________________
> webkit-dev mailing list
> webkit-dev at lists.webkit.org
> http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
>
>


-- 
................................................................

*Sencha*
Jarred Nicholls, Senior Software Architect
@jarrednicholls
<http://twitter.com/jarrednicholls>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-dev/attachments/20111128/05f021ad/attachment.html>