Re: [Gm2] Internal error on simple program

[Gaius Mulley]

Waldek Hebisch <address@hidden> writes:

> The following simple program:
> MODULE tst2;
> PROCEDURE tst1;
> TYPE ind0 = [-100..100];
> CONST low = ind0(-100);
> BEGIN
> END tst1;
> 
> BEGIN
>   tst1
> END tst2
> .
> 
> Couses internal error in gm2:

thanks
> 
> I compiled gm2 from `gm2-cvs-20031013.tar.gz' snapshot. 
> By the way, the following variant:
> MODULE tst5;
> TYPE ind0 = [-100..100];
> CONST low = ind0(-100);
> CONST high = ind0(100);
> TYPE ind = [low..high];
> VAR a : ARRAY [0..100] OF INTEGER;
> VAR b : ARRAY ind OF INTEGER;
> BEGIN
> END tst5
> .
> 
> is rejected by gm2 with error message:
> tst5.mod:5:circular dependancy found when trying to resolve symbol ind
> 
> Note that both programs were actually attempts to break the compiler.
> I wanted array indices to have lower precision then machine words. I
> have noticed that gm2 promotes variables (in `default_conversion'), but
> leaves constants alone, so I have tried to make constants of lower 
> precision.
>  

many thanks for the bug reports, the first test program (tst2) is
currently under investigation..

Gaius