Re: [Confuse-devel] Re: [PATCH 1/7] confuse: ISO C90 does not support the ???%lf??? printf format

[Martin Hedenfalk]

12 jul 2010 kl. 08.40 skrev Carlo Marcelo Arenas Belon:

> On Sat, Mar 27, 2010 at 03:08:28AM -0700, Carlo Marcelo Arenas Belon wrote:
>> 
>> '%f' is valid for float and double, while '%Lf' is needed for long double
>> leaving this as a undefined behaviour for C90 and ignored in C99 as shown
>> by :

We're not using any long doubles, so any %lf formats should just drop the 'l' 
modifier.


>>  confuse.c:1410: warning: ISO C90 does not support the ???%lf??? printf 
>> format
> 
> this will break, at least in mingw32 when using -std=c89 as shown by (not
> confuse specific but still a good example of what would happen with
> cfg_opt_nprint_*)
> 
>  $ cat > float.c 
>  #include <stdio.h>
> 
>  int main(int argc, char *argv[])
>  {
>    float f = 3.1415926;
>    double d = 3.1415926;
> 
>    printf("f = %f, d = %lf\n", f, d);
>    return 0;
>  }^D
>  $ i686-pc-mingw32-gcc -std=c89 -O2 -g -pipe -Wp,-D_FORTIFY_SOURCE=2 
> -fexceptions --param=ssp-buffer-size=4 -mms-bitfields float.c -o float.exe
>  $ ./float.exe 
>  f = 3.141593, d = 0.000000
> 
>> diff --git a/src/confuse.c b/src/confuse.c
>> index 003d0b9..6e7a0ef 100644
>> --- a/src/confuse.c
>> +++ b/src/confuse.c
>> @@ -1407,7 +1407,7 @@ DLLIMPORT void cfg_opt_nprint_var(cfg_opt_t *opt, 
>> unsigned int index, FILE *fp)
>>             fprintf(fp, "%ld", cfg_opt_getnint(opt, index));
>>             break;
>>         case CFGT_FLOAT:
>> -            fprintf(fp, "%lf", cfg_opt_getnfloat(opt, index));
>> +            fprintf(fp, "%f", cfg_opt_getnfloat(opt, index));
>>             break;
>>         case CFGT_STR:
>>             str = cfg_opt_getnstr(opt, index);
>> -- 
> 
> this snippet should prevent that, but wasn't committed as part of this patch
> in 6d5335a8ac0ea6bdb553931a29bef0e2d21d4e3e
> 
> anyone has any objections or is depending on the undefined behaviour that
> "%lf" has?

We don't want any undefined behaviour. Ok.

        -martin