Re: [Gm2] Problem in dtoa?

[Gaius Mulley]

Michael Lambert <address@hidden> writes:

> I'm wondering if there is an issue with dtoa (at least on 32-bit
> machines).  Below is the output of debugging with print statements in
> realconv.mod and the ISO module RealConv.mod.
>
> The 'point = n' comes from the value returned in the call to dtoa
> within the LengthFixedReal procedure.  The 'Length... = n' is from a
> dump of all the calls inside the realconv.mod test.  It looks to me as
> though the first call should have point=2 and the second point=3.
>
> Michael
>
>
> $ ./realconv.x0
> LengthFixedReal(12.3456789, 3) = point = +1
> 5
> LengthFixedReal(123.456789, 3) = point = +2
> 6
> LengthFixedReal(1234.56789, 3) = point = +4
> 8
> LengthFixedReal(1234.56789, -3) = point = +4
> 6
> point = +1/home/lambert/gnu/gcc-4.1.2/gcc/testsuite/gm2/isolib/run/
> pass/realconv.mod:48:in Assert has caused assert failed

Hi Michael,

ahh thanks for this information.  I'm in the process of writing some
runtime regression tests for REAL <-> string conversion (numerical
examples taken from the ISO standard).  Hopefully they will flush out
the bugs when run one different architectures,

regards,
Gaius