Re: What is the best way for termination analyzers to handle pathological inputs?

On 6/11/25 10:18 AM, olcott wrote:

On 6/11/2025 3:42 AM, Mikko wrote:

On 2025-06-10 17:14:58 +0000, olcott said:
>

On 6/10/2025 2:33 AM, Mikko wrote:

On 2025-06-09 21:14:58 +0000, olcott said:
>

The official "received view" of this is that the best
we can possibly do is to do nothing and give up.

>
There is no official view about "the best". What is the best
depends on what one needs and wants. Some may think that the
best they can do is to waste their life in trying to do the
impossible.

>
Certainly doiing nothing and giving up
are not the best thing to do in this case.

>

 void DDD()
{
   HHH(DDD);
   return;
}
 We could just go by the easily verified fact that DDD
correctly simulated by HHH cannot possibly reach its
own "return" instruction final halt state conclusively
proving that the input to HHH(DDD) specifies non-halting
behavior.

Except that only applies WHEN HHH correctly simulates its input, which if it does, it doesn't answer,
And to make that claim, you need to retract your claim that neither HHH nor DDD are actually programs, as both need to be to make the claim.

 int main()
{
   DDD(); // calls HHH(DDD)
}
 The whole idea that HHH(DDD) must report on the behavior
of its caller is just nutty.
 

It is at least a valid way. Another valid way is to run forever, i.e.
until the user gives up and aborts the analyzer.
>
For a particular purpose something else could be better if it does
not cost too much more.
>
Often it is useful to have an analysis of the problem so that the
user of the analyzer can continue the analysis with other tools
or change the program so that the analyzer can analyze more of it.
>
It is not necessary that the analyzer identifies the input as
"pathological" but that identification may be useful in some cases.
>