On Fri, 2013-04-26 at 17:47 +0100, Brian Drummond wrote:
> On Fri, 2013-04-26 at 17:41 +0200, Jean-Christophe Le Lann wrote:
> > Hi all !
> > 
> > I hit a bug in an old version of GHDL :
> > GHDL 0.29 (20100109) [Sokcho edition]
> > Compiled with GNAT Version: 4.6

> >    constant N : natural := 8;
> >      variable n : natural := 0;
> >        n := (n+1) rem N; -- SIGFPE error.

It puzzled me why the error produced was an FPE for a while. 
However given the initial value of n=0, this is equivalent to 
n := (expr) rem 0; 
which is a common or garden divide by zero error.

And cross-checking with Xilinx ISIM, it reports:

ERROR: at 0 ps: rem 0 asked for.  Divide by zero error.
ERROR: In process testcase.vhd:stimuli 

Therefore ghdl's behaviour is entirely correct : the most we could ask
is to improve the runtime error message from "Floating point exception" 
to hint at a possible (probable?) divide by zero error.

- Brian


_______________________________________________
Ghdl-discuss mailing list
[email protected]
https://mail.gna.org/listinfo/ghdl-discuss

Reply via email to