[illumos-Developer] Ugh... xdr_{float,double} problems

[Garrett D'Amore]

It looks reasonable to me.

	- Garrett

On Thu, 2011-02-10 at 21:29 -0600, Jason King wrote:
> To hopefully close this issue (finally).  Here is an updated webrev:
> http://cr.illumos.org/view/nd9d14o6/
> 
> I'm running nightly on an merged workspace at the moment (it takes
> about 3-4 hours for sparc, so I won't be able to validate until
> tomorrow evening after work, but the changes from the last time I ran
> nightly are largely cosmetic, so I don't anticipate any issues).  The
> test case is attached, though there is an outstanding issue with
> NaN/-NaN as it seems x86 & sparc aren't representing them the same --
> specifically they seem to be flipped (but that seems to be an issue
> independent of this change, and something that would take more
> investigation to understand what's going on in the wonderful world of
> floating point).
> 
> If no objections, tomorrow I'll upload (somewhere) the nightly output
> + test results + hg {outgoing -v, pbchk} for a ready
> vict^H^H^H^Hadvocate :)
> 
> On Tue, Feb 1, 2011 at 1:32 PM, Deano <deano at rattie.demon.co.uk> wrote:
> >
> > Garrett Wrote
> >> Not disagreeing with that point, but don't we have the potential to have
> >> a reasonable "portable" version here that works for all platforms
> >> (albeit slower)?
> >
> > If the portable version is, then that is fine but I got the feeling (maybe
> > incorrectly) that this is one of these cases where it is "probably" portable
> > but might have obscure corner cases. In that case I'd rather have an compile
> > warning "This code should work but buyer beware" when working on the port,
> > least then you know and can take appropriate action.
> >
> >> Having such a portable version might also aid people who want to develop
> >> a CPU specific optimized version, since then they'll have a known good
> >>implementation to compare against.
> >
> > Agreed. In situations like this, I'd leave the portable version in even with
> > an #error, its useful even if not perfect on a particular platform as a
> > reference.
> >
> > Bye,
> > Deano
> > deano at cloudpixies.com
> >
> >
> > _______________________________________________
> > Developer mailing list
> > Developer at lists.illumos.org
> > http://lists.illumos.org/m/listinfo/developer
> >
> _______________________________________________
> Developer mailing list
> Developer at lists.illumos.org
> http://lists.illumos.org/m/listinfo/developer