🚀 go-pugleaf

rca@cs.brown.edu (Ronald C. Antony) writes:

>Well, not to protect NeXT in regards to sleeping over important bug
>reports (sometimes I also wonder how they set the priorities, altough
>certainly nobody could claim they are otherwise lazy), I still wonder
>if they are not right. Maybe there are just two opinions what ANSI-C
>is supposed to do.

>The C-Bible "The C Programming Language" by K+R, 2nd edition, ANSI-C
>states the following on page 244:

>g,G	double; %e or %E is used if the exponent is less than -4 or
>	greater than or equal to the precision; otherwise %f is used.
>	Trailing zeros and a trailing decimal point are not printed.


>Now if they got it right, NeXT got it right. If K+R got it wrong,
>well, then NeXT should correct the bug and tell Prentice Hall to
>correct their bug as well.

Well, you cannot have two ways of defining standard and The Standard
says explicitely about g, G conversions "... the precision specyfying
the number of significant digits".  K&R/2 chopped 7 lines of a
description from The Standard into 3 lines dropping an essential
piece of an information in the process.

What is more I know of many computers around which got it right.
Look for example at gawk sources and find out how many of them
need -DGFMT_WORKAROUND to alleviate problems of buggy libraries.
Library correctness does not seem to be very high on NeXT list.
BTW - I was also reporting this problem in early days of 2.1.
Everybody can see results. :-)

    Michal Jaegermann
    ntomczak@vega.math.ualberta.ca