[mercury-users] Self-referential data structures

[Tyson Dowd]

On 26-Nov-1997, Richard A. O'Keefe <ok at cs.rmit.edu.au> wrote:
> 
> If an operation doesn't make sense, then Mercury should produce an
> _intelligible_ error message when it happens.  
> 	"foo/5;2: Attempt to get value stored in a null mutvar"
> is a good response.
> 	"Segmentation fault, core dumped."
> is not.

I agree. Doing the checks required to give such a message might cause
performance problems for this code. I haven't looked at this case very
carefully, but it might be possible to implement good error messages
along with good efficiency using a virtual memory hack. If, instead of
null, a reasonable pointer value is stored that happens to point into a
protected page (no read, no write), then the signal handler can catch
this message and will be able to tell by the faulting address that it
was a "uninitialized mutvar" error. But this might cause problems with
the GC.

The foo/5;2: part of your error message isn't so easy, but it just
so happens I've been working on stack frame layout data structures
that should allow us to give a stack trace after any runtime error
(stack overflow, out of memory, etc).

-- 
       Tyson Dowd           # 
                            #         Linux versus Windows is a 
     trd at cs.mu.oz.au        #            Win lose situation.
http://www.cs.mu.oz.au/~trd #