Integration history file does not contain a graphics.

[Juergen Reuter]

Description says it. Investigating.

[Juergen Reuter]

Trivial example is

model = SM
process bla = e1, E1  => e2, E2
compile
sqrts = 500 GeV
integrate (bla)

which generates an arithmetic exception in the metapost code (don't know why yet):

Package: `gamelan' 0.40 <1998/11/02>
! Arithmetic overflow.
Gfindtransf_->...ormed(SUFFIX1)=urcorner.theframe;
                                                  if.Gtype_.x<0:(SUFFIX1):=(...
<to be read again>
                   ;
Gfindconv_->...(SUFFIX5);Gfindtransf_((SUFFIX3))t;
                                                  Gdefconv_(t);endgroup
freeze->...d_:Gfindconv_(Gfixed_,Gr_,Glsc_,Glmax_)
                                                  ;Gfixed_:=true;for.i=1upto...

hide->exitif.numeric.begingroup(TEXT0)
                                      ;endgroup;
auto->hide(freeze)
                  hide(Fe_base:=Fen_base)hide(def.Gcma_=hide(def.Gcma_=,endd...
<to be read again>
                   :
<for(1)> ...oc_=auto.if.labxf.lft=0.5:x.else:y.fi:
                                                  hide(Gmks_:=Gmks_+1)hide(G...
<for(lft)> ...)(Gtform_[c_],loc_))endfor;fi.endfor
                                                   ENDFOR
grid->...Gtform_[c_],loc_))endfor;fi.endfor.endfor
                                                  .Greset_ticks_;Gneedgr_:=f...
<to be read again>
                   ;
fullgrid->...n_[1]noticks.urt,n_[2]iticks)(TEXT4);
                                                  endgroup
<to be read again>
                   ;
l.74   fullgrid.lr (5,20);

[Juergen Reuter]

The arithmetic exception happens in the GAMELAN function graphrange, but I couldn't trace it down yet.

[Juergen Reuter]

The problem is when the graphrange is too small, e.g. 427 and 434, then graphrange runs into an arithmetic exception.

[Juergen Reuter]

WK, could you find a solution for that, please... I think it is the determination of the scale factor in Gfindconv...

[kilian]

I'm not too confident that this can be fixed. The numeric range of Metapost parameters is very limited, and therefore overflows are frequent.

However, I'll check this. After all, it is in the grid code which merely draws axis tick marks and labels.

[kilian]

I've not dared to delve into the Gamelan code, but a sensible workaround is to limit the graph range. Done in r3183.