[PATCH] c/67882 - improve -Warray-bounds for invalid offsetof

[Martin Sebor]

On 10/23/2015 11:45 AM, Bernd Schmidt wrote:
> On 10/23/2015 06:50 PM, Joseph Myers wrote:
>> On Fri, 23 Oct 2015, Martin Sebor wrote:
>>
>>> But now that I'm re-reading the answer above I see that Joseph
>>> was suggesting that a5_7[5][0] should be diagnosed when the patch
>>> accepts it as an extension.  I think we do want to accept it
>>> because a5_7 is treated as a flexible array member (as an extension)
>>> and so the upper bound of the major index is unknown. I.e., FA5_7
>>> is defined like so:
>>
>> If you treat it as a flexible array member, then, yes, it would be valid.
>
> Ok, let's install the patch as-is, and postpone the discussion of
> whether that is a valid flexible array member (I certainly wouldn't have
> guessed so from the documentation which only mentions [], [0] and [1] as
> valid cases).

The original code deliberately avoids diagnosing the case of last
array members with bounds greater than 1 (see the comment about
"a poor man's flexible array member" added with a fix for bug
41935) and I didn't want to change that.

But if there is sentiment for tightening it up I would be very
much in favor. IMO, it would be ideal if we could agree on and
apply the same rules for offsetof as for other expressions (and
diagnose, for example, &a5_7[5][0], which currently isn't
diagnosed).

As I mentioned, I'm planning to work on bug 67872 and it would
be helpful to know what our rules are up front. I can go back
and update this patch after it's been committed if the rules
evolve between now and then.

>
> I guess this is a case where I could say either "I wrote the patch" or
> "I requested changes to a patch in review"; in the latter case I can
> approve it. Joseph seems on board with what we've discussed, so I'd say
> please wait until Tuesday for objections then commit.

Okay.

Martin