[PATCH] Add warnings for undefined behaviors with pointers

[Richard Smith]

On Fri, Jun 6, 2014 at 4:51 PM, Richard Trieu <rtrieu at google.com> wrote:

> On Fri, Jun 6, 2014 at 3:58 PM, Richard Smith <richard at metafoo.co.uk>
> wrote:
>
>> On Thu, Jun 5, 2014 at 5:01 PM, Richard Smith <richard at metafoo.co.uk>
>> wrote:
>>
>>> LGTM with some minor diagnostic tweaks.
>>>
>>> ================
>>> Comment at: include/clang/Basic/DiagnosticSemaKinds.td:2393
>>> @@ +2392,3 @@
>>> +  "'this' pointer cannot be null in well-defined C++ code; pointer may
>>> be "
>>> +  "assumed always converted to a true value">,
>>> InGroup<UndefinedBoolConversion>;
>>> +def warn_address_of_reference_bool_conversion : Warning<
>>> ----------------
>>> "assumed to always convert to" would sound better to me.
>>>
>>
>> Looks like you didn't make this change prior to checkin?
>>
>
> I'll go fix it.
>

Thanks!


>  Also,
>>>
>>>   assumed to always convert to 'true'
>>>
>>> rather than
>>>
>>>   assumed to always convert to a true value
>>>
>>> to avoid people wondering what "a true value" is (cf. "a true Scotsman").
>>>
>>> Likewise I'd suggest putting the %select's below into single quotes.
>>>
>>
>> Likewise this one.
>>
>
> Currently, there's only one place in the file that uses single quotes
> while there are several that don't use quotes.  For consistency, I left the
> quotes off.
>

OK =)


>
>
>> ================
>>> Comment at: include/clang/Basic/DiagnosticSemaKinds.td:2409
>>> @@ +2408,3 @@
>>> +  "reference cannot be bound to dereferenced null pointer in
>>> well-defined C++ "
>>> +  "code; comparison maybe be assumed to always evaluate to "
>>> +  "%select{true|false}0">,
>>> ----------------
>>> Typo "maybe be" should be "may be"
>>>
>>> http://reviews.llvm.org/D3999
>>>
>>>
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20140606/2db090fe/attachment.html>