Re: address warnings from GCC's UB sanitizer

[Akim Demaille]

> Le 15 mars 2019 à 18:28, Akim Demaille <address@hidden> a écrit :
> 
> Hi Paul,
> 
>> Le 14 mars 2019 à 22:26, Paul Eggert <address@hidden> a écrit :
>> 
>> On 3/14/19 1:43 PM, Akim Demaille wrote:
>> 
>>> +  return column <= INT_MAX - width ? column + width : INT_MAX;
>> 
>> Nowadays a better (i.e., more-efficient, and I think easier-to-follow)
>> way of writing this is:
>> 
>>  int result;
>>  return INT_ADD_WRAPV (column, width, &result) : INT_MAX : result;
>> 
>> where INT_ADD_WRAPV is defined in intprops.h.
> 
> Excellent!  Thanks a lot.  I'll install the following when the CI confirms it.

GCC 4.9, 4.7 and 4.6 generate warnings about this.

For instance https://travis-ci.org/akimd/bison/jobs/506935039

src/location.c: In function 'add_column_width':
./lib/intprops.h:430:50: error: comparison is always false due to limited range 
of data type [-Werror=type-limits]
     || (EXPR_SIGNED ((a) op (b)) && ((a) op (b)) < (tmin)) \
                                                  ^

And Clang 3.3, 3.4.  For instance 
https://travis-ci.org/akimd/bison/jobs/506935052

src/location.c:45:10: error: comparison of constant -9223372036854775808 with 
expression of type 'int' is always false 
[-Werror,-Wtautological-constant-out-of-range-compare]  CC       src/bison-lr0.o
  return INT_ADD_WRAPV (column, width, &res) ? INT_MAX : res;
         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~


The whole result is here: https://travis-ci.org/akimd/bison/builds/506935032.

Are you aware of any way to use this macro and avoid these spurious warnings?