Re: DDD incorrectly emulated by HHH is INCorrectly rejected as non-halting V2 --- key point needing review

On 7/15/24 12:23 PM, olcott wrote:

On 7/15/2024 11:03 AM, Mike Terry wrote:

On 15/07/2024 09:59, joes wrote:

Am Sun, 14 Jul 2024 22:35:03 -0500 schrieb olcott:

On 7/14/2024 10:02 PM, Mike Terry wrote:

On 15/07/2024 01:20, joes wrote:

Am Sun, 14 Jul 2024 09:00:55 -0500 schrieb olcott:

On 7/14/2024 3:29 AM, joes wrote:

Am Sat, 13 Jul 2024 18:33:53 -0500 schrieb olcott:

On 7/13/2024 6:26 PM, joes wrote:

>

[0000217a][0015e2dc][0000217f] e853f4ffff call 000015d2 ; call
HHH(DDD) Local Halt Decider: Infinite Recursion Detected Simulation
Stopped

How is this detected?

PO seems not to want to answer you, as I notice you've asked this
question more than once and PO dodges a direct response, so I'll try.
(Alternatively, PO has provided a link to his source code in the past,
so if you can find that link you can just look the answer yourself -
the functions are all in his halt7.c file, which is compiled but not
linked, then the obj file is interpreted within his x86utm.exe (source
also given in the link.  The link might not reflect his current code??)

Thank you. I didn't bother digging through their code, and they refused
to give the abortion criterion.
>

HHH [outer HHH only!] examines a global trace table of simulated
instruction (from all simulation levels merged together).  The
particular message "Infinite Recursion Detected Simulation Stopped"
seems to be issued when:
-  last instruction is a CALL -  working backwards through the merged
trace table, another CALL is encountered -  ..which is issued at the
same address -  ..and is calling to the same address -  ..and no
"conditional branch" instructions occur in the trace table
       between the two call instructions
>
KEY TO NOT BEING MISLED BY THE ABOVE:
>
0. The "Infinite Recursion Detected Simulation Stopped" message is just
a printf.
     It does not prove that /actual/ infinite recursion was detected -
     on the contrary,
     all here but PO realise that the recursion detected is just
     finite recursion.
>
1. The trace table being examined is NOT an x86 processor trace - it is
a "merged simulation trace" containing entries for ALL SIMULATION
     LEVELS.
     So the two CALL instructions are not referring to one single x86
processor.

When emulated DDD calls HHH(DDD) the outer HHH emulates itself emulating
DDD.
I think that joes does not understand these things.

>

     Typically, the last call instruction is from a deeper nested
     simulation than the earlier detected call instruction.  The outer
     simulations are all
     still running, but do not appear in the trace table or logs
presented by PO due to the next note.
>
2. The searched trace table is filtered to only contain instructions
within the C function D/DD/DDD/.. !!
     YES, YOU READ THAT RIGHT!  ALL CODE IN HHH IS TOTALLY IGNORED,
INCLUDING
     THE CONDITIONAL BRANCH INSTRUCTIONS THAT ARE TESTING THE VERY
     ABORT TESTS THAT CAUSE OUTER HHH TO ABORT.
>
3. Inner HHH's do not perform the same tests as above, because they
inspect a global
     variable which tells them they are inner HHH's.  Yeah, that means
the simulation
     is completely broken logically... [but... the outer HHH will
     abort first, so
     PO might argue the outcome will be the same, even though
     logically it is broken...]

Ah, and here I believed them when they said they had rewritten it.

>
I doubt he has done it properly if at all.  I.e. I'm confident there will still be mutable static data giving different code paths for outer/inner executions.
>

 That is counter-factual.
 

I offered to help PO sort out the required recursive logic but this was ignored, PO saying it would take him thousands of years(?) to do it correctly.  I reckon anyone else here would fix it in a couple of hours!
>

 *THIS IS THE KEY POINT NEEDING MIKES REVIEW*
I propose that the use of static local variables to
pass information up from the slaves to the master
does not violate computability.
 

But the inner one CAN NOT KNOW it is an inner one without violating the constraint.
Now, all this seems to be an attempt to make it so the decider can know that it is seeing a copy of itself, but once it knows that, then the necessary condition that whatever the outer one does must be compatible with what it has decided the inner one will do applies, or you are just admitting that it did incorrect logic on the determination of the behavior of the inner one. Remember also, just because the outer simulator aborts the inner doesn't mean the behavior of the program represented by the input stopped progressing, as ALL copies of it do the same thing, and continue in their behavior till they reach their end.
And of course, the REAL problem is figuring out how every HHH can return to main but not to DDD, which means that it CAN'T be a pure function, as that violates the requirement. This is because BY DEFINITION, the "behavior of the input" is the actual behavior of the program that it represents, and NOT the results of the PARTIAL emulation done by the HHH that returned the answer.