[ovs-dev] "array subscript is above array bounds" for flowmap_set(), flowmap_clear()

[Jarno Rajahalme]

> On Aug 26, 2015, at 4:50 PM, Ben Pfaff <blp at nicira.com> wrote:
> 
> GCC 4.9 doesn't like flowmap_clear() and flowmap_set().  Maybe that's
> why you had those seemingly redundant tests in there?  I'm taking a
> closer look to see whether I believe its claims though.
> 
>      CC       lib/classifier.lo
>    In file included from ../lib/match.h:20:0,
>                     from ../lib/classifier.h:302,
>                     from ../lib/classifier.c:18:
>    ../lib/classifier.c: In function 'miniflow_get_map_in_range':
>    ../lib/flow.h:520:28: error: array subscript is above array bounds [-Werror=array-bounds]
>             fm->bits[unit + 1] &= ~(n_bits_mask >> (MAP_T_BITS - idx));
>                                ^
>    ../lib/flow.h:520:28: error: array subscript is above array bounds [-Werror=array-bounds]
>    cc1: all warnings being treated as errors
>    Makefile:3910: recipe for target 'lib/classifier.lo' failed
>    make[2]: *** [lib/classifier.lo] Error 1
>      CC       lib/flow.lo
>    In file included from ../lib/flow.c:18:0:
>    ../lib/flow.c: In function 'miniflow_extract':
>    ../lib/flow.h:502:28: error: array subscript is above array bounds [-Werror=array-bounds]
>             fm->bits[unit + 1] |= n_bits_mask >> (MAP_T_BITS - idx);
>                                ^
>    ../lib/flow.h:502:28: error: array subscript is above array bounds [-Werror=array-bounds]
>    cc1: all warnings being treated as errors

This rings a bell, and today I was compiling with clang only, so I did not see them.

Maybe we introduce the tests back with a comment stating the reason we have them. They will be removed by the compiler anyway, as both relevant arguments (‘idx’ and ’n_bits’) are compile-time constants.

  Jarno