Re: [bug-gawk] option --non-decimal-data discard lowest 8 bits

[arnold]

Hi.

Thanks for submitting a report.

As you surmise, this is a (documented!) limitation of awk's use of double
precision floating point numbers.

You can use the -M option to supply more bits of precision, if your
version of gawk was compiled with MPFR support.

Thanks,

Arnold
Zhizhou Zhang <address@hidden> wrote:

> Hi,
>
> I'm using gawk 4.1.4. I found a problem with --non-decimal-data. It's
> shown below:
>
> $ cat a
> 0x1111222233334444
> $ cat a | gawk --non-decimal-data '{printf( "0x%x\n", $0)}'
> 0x1111222233334400
>
> As it is, the lowest 8 bits 0x44 is discarded. I had a glance over the
> source code, It seems because of AWKNUM is float type, which reserved
> some bits for exponent part.
>
> #define AWKNUM  double
>
> Is this a known issue? I'm not sure whether it is a bug, I hope I
> haven't bothered you too much.
>
> -- 
> Regards,
> Zhizhou