[tarantool-patches] Re: [PATCH 12/13] sql: fixes an error in uint64 to double casting

["n.pettik" <korablev@tarantool.org>]

Merge with one of previous patches.

> On 15 Mar 2019, at 18:45, Stanislav Zudin <szudin@tarantool.org> wrote:
> 
> ---
> src/box/sql/util.c    | 2 +-
> src/box/sql/vdbeInt.h | 2 ++
> src/box/sql/vdbemem.c | 2 +-
> 3 files changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/src/box/sql/util.c b/src/box/sql/util.c
> index 17268aaaa..d585dc0d5 100644
> --- a/src/box/sql/util.c
> +++ b/src/box/sql/util.c
> @@ -1318,7 +1318,7 @@ sqlAddInt64(i64 * pA, bool is_signedA, i64 iB, bool is_signedB)
> 			if (sum == INT64_MIN_MOD) {
> 				*pA = INT64_MIN;
> 			} else {
> -				assert(sum < INT64_MAX);
> +				assert(sum <= INT64_MAX);
> 				*pA = -(i64)sum;
> 			}
> 			return ATHR_SIGNED;
> diff --git a/src/box/sql/vdbeInt.h b/src/box/sql/vdbeInt.h
> index 42f22df52..2076a9a14 100644
> --- a/src/box/sql/vdbeInt.h
> +++ b/src/box/sql/vdbeInt.h
> @@ -251,6 +251,8 @@ struct Mem {
> #define MEM_Unsigned  0x20000	/* Value is unsigned integer.
> 				 * Combine this flag with MEM_Int
> 				 * if necessary */
> +#define MEM_UInt (MEM_Int | MEM_Unsigned)
> +

IMHO it is a terrible mess. Don’t use such macros.
If MEM_Unsigned can serve only as an attribute of
MEM_Int, then you could declare its value already
containing necessary bit.