gnuplot-bugs

[Gnuplot-bugs] segmentation fault

[Jan S. <jsc...@un...>]

Hello,

i had a segmentation fault due to long recursion of a function.
Actually that was an programming error on my side but this should
not result in a segmentation fault.

I could reproduce the problem with this function:

iter( x , i ) =3D (i =3D=3D 0) ? x : iter( x/2. , i-1 )

calling iter(.) with a noninteger i makes the i=3D=3D0 test never true ..=
=2E



------------------------------------------------------------
   Dr. Jan H. Schween
   Inst. f. Geophysics and Meteorology
   Z=FClpicher Str. 49a
50674 Koeln
   Germany
+49 (0)221 470 17 76 (fax ... 5161)
jsc...@un...
------------------------------------------------------------

Re: [Gnuplot-bugs] segmentation fault

[<HBB...@t-...>]

Jan Schween wrote:

> i had a segmentation fault due to long recursion of a function.
> Actually that was an programming error on my side but this should
> not result in a segmentation fault.

There's not much choice in the matter.  This segmentation fault is 
caused by stack overflow, which is impossible to predict in any portable 
fashion.  We would have to introduce an arbitrary limitation of call 
stack depth, and for that limit to work, it would have to be decided by 
the smallest stack size on any platform running gnuplot.  I.e. we would 
deliberately break scripts although they could still work on platforms 
with bigger stacks, for the sole benefit of ill-written scripts.

The last time this issue came up, we decided to leave it as-is.