Hercules crash at dyn_perform_random_number_operation

[salva-rczero]

For some execution of PRNO, function code 114 (PRNO-TRNG), Hercules crashed due to an invalid memory address.

After some testing, I traced it to these lines:

randbytes = regs->GR( r2 + 1 );
...
randbytes = regs->GR( r1 + 1 );
...
randbytes = regs->GR( r1 + 1 );
...
randbytes = regs->GR( r2 + 1 ); 

Here, randbytes was loading an unexpectedly large value.
I believe (and the POP confirms) that the length field must be a 32-bit number for AMODE 24/31.

So I changed them to:

 randbytes = GR_A(r1 + 1, regs); 
or
 randbytes = GR_A(r2 + 1, regs); 

And It works!

Please consider applying this change. I should provide a patch/pull request, but since it’s a small fix and I don’t quite remember how to, I’m just reporting it here.

Thanks, salva.

[Fish-Git]

Please see: SUBMITTING PROBLEM REPORTS

Knowing what version of Hercules you are using as well as knowing what instruction sequence was used (i.e. a reproducible test case for us to try for ourselves) as well we what Hercules host you are running on as well as the Crash Dump itself that could then be analyzed, etc, would go a LONG WAY with helping us to help you.

As it stands right now, you are asking us to "please make this change to fix my problem" without providing any convincing evidence that your suggested change is valid and worth our time.

Sorry, but we don't work that way.

Please try again.

[salva-rczero]

No, I'm not saying "please make this change to fix my problem". I say: I found a bug and solved my problem making these changes. I share it with you for your consideration.

If this somehow isn’t respectful enough, or if I’m breaking the holy handbook of how one should dare address the almighty guardian of Hercules, then please forgive my reckless audacity—and feel free to smite me with your infinite wisdom.

Regards, salva.

[jordan31bit]

Well, at least they recognize your authority Mr. Fish.

If this somehow isn’t respectful enough, or if I’m breaking the holy handbook of how one should dare address the almighty guardian of Hercules, then please forgive my reckless audacity—and feel free to smite me with your infinite wisdom.

Regards, salva.

[Fish-Git]

That was totally uncalled for, salva.

All I asked you for was more detailed information in your problem report. All I did was remind you that we Hercules developers are not psychic. We don't know the particulars (specifics) of your environment that caused Hercules to misbehave as it did for you. So I directed you to a GitHub Issue that explains the types of detailed information we typically need when problems are reported, and asked that you please supply it.

Was it rude of me to do that? I did not intend to be disrespectful towards you in any way! You are a valuable contributor to Hercules and I would never be purposely disrespectful towards you! I apologize sincerely if I somehow was!

Are you going to provide the requested information or not?

[salva-rczero]

Fish, I do not have a problem, I have a solution!

Use it, or wait until the bug bites someone more obedient to your rules than me.

[misterspock1]

@salva-rczero Just curious, what would one have to be doing to reproduce a crash? Do you have a test case so I could see the crash in action?

[Fish-Git]

Fish, I do not have a problem, I have a solution!

You have a problem. Two of them in fact:

1. One does not provide "solutions" to non-problems. In order to provide a solution, one first has to have a problem. You even mentioned what the problem was: "Hercules crashed due to an invalid memory address."

2. You are being combative and failing to cooperate (which I really do not understand).

You are the first person I have encountered that has failed to provide additional information when asked for it. Why are you being like this, salva?

I appreciate your "solution", but like Michael, I would like to try and reproduce your original problem for myself as well. I would like to see for myself where this Hercules crash occurs.

I would also like to see where in the Principles of Operation manual where it mentions "the length(?) field must be a 32-bit number". I am not seeing that. Perhaps I am looking in to wrong place? Do you have a page number and paragraph?

Why are you being so obstinate?

[salva-rczero]

@salva-rczero Just curious, what would one have to be doing to reproduce a crash? Do you have a test case so I could see the crash in action?

Load garbage into high word of length operand. E.g. (from tests/PRNO.asm):

         LGFI  R0,114         R0->function code 114: TRNG
         LA    R2,FO          R2->first  operand address
         LGFI  R3,64          R3->first  operand length
         LA    R4,SO          R2->second operand address
*         LGFI  R5,64          R3->second operand length
         LG     R5,=x'0505050500000040'
         PRNO  R2,R4          perform random number generate operation

My fix for AMODE24/31 solved my problem, but Hercules in AMODE64 will still crash if an invalid length is specified.

[salva-rczero]

I would also like to see where in the Principles of Operation manual where it mentions "the length(?) field must be a 32-bit number". I am not seeing that. Perhaps I am looking in to wrong place? Do you have a page number and paragraph?

What is yor POP version?
What is the software you are using to open it? Acrobat Reader?
What is your O.S.? Windows? Linux?

I'm not a psychic. Please, try again.

[wrljet]

I would also like to see where in the Principles of Operation manual where it mentions "the length(?) field must be a 32-bit number". I am not seeing that. Perhaps I am looking in to wrong place? Do you have a page number and paragraph?

What is yor POP version? What is the software you are using to open it? Acrobat Reader? What is your O.S.? Windows? Linux?

I'm not a psychic. Please, try again.

OK, that's just about enough.

[Fish-Git]

OK, that's just about enough.

I'm close to reaching my limit too, Bill.  ):-<

I've already apologized to him, but apparently he refuses to accept it.

[Fish-Git]

What is yor POP version?

SA22-7832-14

What is the software you are using to open it? Acrobat Reader?

PDFXChange Viewer version 2.5 (build 322.10).

What is your O.S.? Windows? Linux?

Windows 7 x64 SP1.