bug#15307: minor fix to dfa.c

[Jim Meyering]

On Sun, Sep 8, 2013 at 11:01 AM, Jim Meyering <address@hidden> wrote:
> On Sun, Sep 8, 2013 at 2:53 AM, Aharon Robbins <address@hidden> wrote:
>> The following fix to dfa.c was suggested by a static checking tool.
>> I'm applying it in the gawk code base.
>>
>> Basically, it's theoretically possible for len to have run off the end
>> of the `str' array.
>>...
>
> Hi Arnold,
>
> Thanks, but that makes it look like "str" will instead fail to be
> NUL-terminated,
> in which case the following strcmp (aka STREQ) would overrun the buffer.
> Yes, this is all theoretical, but still...
>
> I see that the current limit is 31:
>
>   $ for i in 30 31 32 33; do printf "$i "; src/grep -E '[[:'$(perl -e
> 'print "a"x'$i)':]]'; done
>   30 src/grep: Invalid character class name
>   31 src/grep: Invalid character class name
>   32 src/grep: Unmatched [ or [^
>   33 src/grep: Unmatched [ or [^
>
> So I propose this patch instead:

Hi Arnold,

I was going to push that change, but then realized I didn't know
which static analysis tool you were referring to.  Which was it?