Re: Gutenprint 5.3.4: only on 32-bit failures with "Image is too long/wide for the page" for certain PPDs

Hello,

On 2021-09-11 17:51, Robert Krawitz wrote:
> On 9/9/21 7:58 AM, jsmeix wrote:
>> From my own experience about 30 years ago I remember
>> that in practice
>> "there is no such thing as 'equal' with float data types".
>> 
>> I.e. code that is based on 'float1 == float2' cannot work
>> reliably in general because even if float1 and float2
>> must me mathematically identical it does not mean their
>> representations as a limited sequences of bits are same.
> 
> That's certainly true, but the converse (that
> inequality comparisons are safe) is not necessarily
> true (even strict inequality comparisons,
> e. g. < vs. <=, aren't always safe for numbers
> that can't be represented exactly).

Yes, of course.
That's what I meant with "based on 'float1 == float2'".
E.g. 'float1 < float2' is based on 'float1 == float2'
because when one cannot safely/reliably determine
when 'float1 == float2' holds then one also cannot
safely/reliably determine when 'float1 < float2' holds
because 'float1 < float2 <=> 'NOT float1 >= float2' and
'float1 >= float2' <=> 'float1 > float2' OR 'float1 == float2'
so 'float1 < float2' needs (is based on) 'float1 == float2'.


> We may need to implement small fudge factors,
> maybe something like 0.000001 point

I would not call it "fudge" because it is not about
to somehow "cheat" a bit here but the contrary:
It is about to ignore meaningless aritfacts
of differences in some least significant bits.
I.e. ignoring such meaningless aritfacts
makes things working correct.


Kind Regards
Johannes Meixner
-- 
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5 - 90409 Nuernberg - Germany
(HRB 36809, AG Nuernberg) GF: Felix Imendoerffer