Re: Lout 3.17 runs not converging

[Valeriy E. Ushakov]

On Thu, Dec 02, 1999 at 08:27:32AM +0300, Mikko Huhtala wrote:

> > > ... keeps giving these warnings:
> > >     9,1: unresolved cross reference @Runner&&following
> > > no matter how many times I run Lout over the input.
> 
> I have had the same problem on both Red Hat 6.0 and SGI Irix 6.4 and 6.5.

[NB: Matej Cepl complained about the same problem with 3.16]

Could this be a floating point problem?  E.g. a tiny rounding error
forces an object to be pushed back and forth between two pages and so
the xref db is always one run behind the current state of the
document.  Can someone who experience this problem try recompiling
with wahtever flags are necessary to force Intel to do correct FP
math and then see if the problem goes away.

AFAIK, Intel has two FP modes, "strict" and "wide", I'm not sure how
you can ask gcc to use particular one.  Perhaps -ffloat-store could
help as well.

PS: I now recall that 3.16 on win32 was giving me this warning until
the final run (which surprised me), but ediffing its output with
preformatted user.ps revealed only minor FP differences, IIRC.  I'm
not sure if I instructed compiler to use strict math, but I believe so
- I need to check.

SY, Uwe
-- 
address@hidden                         |       Zu Grunde kommen
http://www.ptc.spbu.ru/~uwe/            |       Ist zu Grunde gehen