[cfe-commits] [Patch] Remove false positives from warnings on NULL in arithmetic expressions.

[Chandler Carruth]

On Sun, Jun 19, 2011 at 1:24 AM, Chandler Carruth <chandlerc at google.com>wrote:

> On Fri, Jun 17, 2011 at 7:38 PM, Richard Trieu <rtrieu at google.com> wrote:
>
>> -Wnull-arithmetic was catching some false positives.  Switch the checking
>> code so that it will warn on specific types instead of not warning on
>> specific types.  Now, comparisons between non-boolean arithmetic types and
>> NULL will give this warning.  New types to warn on can be added later if
>> needed.
>
>
> I'm a bit worried that the set of things we *don't* want to warn about is
> more clearly defined than the set of things we *do*. I'm inclined to
> implement the warning accordingly. I think we just need something more
> forgiving than 'isPointerType'. I'm looking at committing a variation of
> this patch (your test cases are excellent) that instead enhances the AST to
> have an interface specifically designed to represent this class of types. We
> can implement that much more efficiently as a bonus, and this is a hot-path
> in Clang.
>

I implemented this idea in r133383. Please review Richard and let me know
your thoughts. Also, if anyone else disagrees with this direction, we can
switch back to Richard's original approach. =] Should be simple code
changes.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20110619/2b97924a/attachment.html>