memset bugs.

Casey Dahlin cdahlin at redhat.com
Mon Nov 30 16:39:32 UTC 2009


On 11/30/2009 10:39 AM, Peter Jones wrote:
> On 11/27/2009 02:25 PM, Casey Dahlin wrote:
>> On 11/27/2009 06:03 AM, Richard W.M. Jones wrote:
>>> On Fri, Nov 27, 2009 at 03:28:19AM -0500, Gregory Maxwell wrote:
>>>> A literal zero prior to preprocessing is either a bug, or some kind
>>>> of dead-
>>>> code causing place-holder.
>>>
>>> Not necessarily .. the C code itself may be generated from
>>> something else.
>>>
>>> Rich.
>>>
>>
>> In which case the C code is no longer "source" and should be excluded
>> from the analysis.
> 
> No, when swig (or whatever) produces bad code, we still want the compiler to
> identify it and toss it.  It's then up to the packager to realize it's swig
> producing the bad code, but it isn't magically good code that doesn't result
> in real bugs.
> 

The compiler isn't doing these checks, but point taken.

On a tangent, what of these checks if any should be put into the compiler? Compile-time bounds checking of library functions is kind of "magical" and un-C-like, but its not unprecedented (printf argument checking for example).

--CJD




More information about the fedora-devel-list mailing list