Re: [PATCH] config.c: fix potential number truncation in git_parse_signed()

[Duy Nguyen <pclouds@gmail.com>]

On Wed, Jul 6, 2016 at 9:33 PM, Junio C Hamano <gitster@pobox.com> wrote:
> Nguyễn Thái Ngọc Duy  <pclouds@gmail.com> writes:
>
>> clang -Wabsolute-value on IA-32 architecture complains that "absolute
>> value function 'labs' given an argument of type 'intmax_t' (aka 'long
>> long') but has parameter of type 'long' which may cause truncation of
>> value". Very unlikely for this code though. Nevertheless, add an
>> explicit check for truncation to shut clang up and error out.
>
> Thanks.  It however makes me wonder if it is a better approach to
> avoid downcasting intmax_t to long by using imaxabs()?

I think the tricky part is detect overflow. WIth uval being a
potentially signed number, i'm not sure how to catch it. The (uval >
max || labs(val) > uval) was written with unsigned number in mind, I
think.

>
>>
>> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
>> ---
>>  config.c | 10 ++++++++--
>>  1 file changed, 8 insertions(+), 2 deletions(-)
>>
>> diff --git a/config.c b/config.c
>> index d7ce34b..880bd4a 100644
>> --- a/config.c
>> +++ b/config.c
>> @@ -503,6 +503,7 @@ static int git_parse_signed(const char *value, intmax_t *ret, intmax_t max)
>>               intmax_t val;
>>               uintmax_t uval;
>>               uintmax_t factor = 1;
>> +             long int lival;
>>
>>               errno = 0;
>>               val = strtoimax(value, &end, 0);
>> @@ -512,9 +513,14 @@ static int git_parse_signed(const char *value, intmax_t *ret, intmax_t max)
>>                       errno = EINVAL;
>>                       return 0;
>>               }
>> -             uval = labs(val);
>> +             lival = (long int)val;
>> +             if (lival != val) {
>> +                     errno = ERANGE;
>> +                     return 0;
>> +             }
>> +             uval = labs(lival);
>>               uval *= factor;
>> -             if (uval > max || labs(val) > uval) {
>> +             if (uval > max || labs(lival) > uval) {
>>                       errno = ERANGE;
>>                       return 0;
>>               }



-- 
Duy