floating point war casualty list

[Toon Moene]

Robert Dewar wrote:

> I wrote:

> > Well, that means that we now have another part of GCC that's not really
> > free software (the other one is reload{1}.c), because it can't be
> > modified.

> I find this an odd summary. Just because one person's attempt to
> modify one piece of software without adequate knowledge was faulty
> is hardly reason to make a blanket statement like this.
> Floating-point stuff is VERY tricky, very few people have sufficient
> knowledge of cross-platform considerations here to make modifications
> in this area safely.

I think you just snipped to much.  My comment was about this remark by
Stephen:

> It is just not a wise move to disturb this stuff,
> because it will take a very long time for the dust to settle down.
> The most generous explanation offered for the recent changes is that
> somebody was on a cosmetic cleanup campaign.  It would be far better
> to let it be, leave it ugly.

I read this as: You'd better not touch this.  He doesn't talk about
capabilities.  He is the original author of this code.  He should have
been able originally to add comments about what considerations went into
its design, which would enable potential modifiers of this code to
assess the difficulties and refrain from updating if they felt not
capable.

> There is no magic in free software that suddenly ensures that anyone
> can change anything successfully :-)

Ture, but there's also no magic in the mathematics of precise floating
point arithmetic that makes it unexplainable :-)

-- 
Toon Moene - mailto:toon@moene.indiv.nluug.nl - phoneto: +31 346 214290
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
Maintainer, GNU Fortran 77: http://gcc.gnu.org/onlinedocs/g77_news.html
Join GNU Fortran 95: http://g95.sourceforge.net/ (under construction)