> On Jan 6, 2019, at 9:34 PM, Fritz Mueller via cctalk <cctalk@classiccmp.org> 
> wrote:
> 
> Hi Paul,
> 
>> On Jan 6, 2019, at 5:58 PM, Paul Koning <paulkon...@comcast.net> wrote:
>> 
>> Hm.  Can you read data back from the RK05 pack?  I'd have to refresh my 
>> memory on how but it's clearly possible to force a crash dump.  That would 
>> allow us to dig into exactly what went wrong, provided you can read the dump 
>> file (or the whole disk).
> 
> Yes, I can read back, either by sector or the whole pack.  Do let me know if 
> you find a way to trigger a crash dump, and I’ll give it a go!
> 
>       thanks,
>         --FritzM.

Here is the procedure.

1. Make sure crash dump is enabled (in the "default" option).  Start the 
system.  Let it run for at least one minute.  (I'm not entirely sure about 
older versions, but I think that a crash within one minute of startup is 
handled differently and doesn't do all the usual dump and restart machinery.)

2. Set the data switches all UP.  (In SIMH, enter "D SR 177777".)

3. Set a breakpoint.

4. When you hit the breakpoint, change the PC to 52, like this:

        0B:055244 
        _$7/055244 52 
        _P 

   (you enter only "$7/" and "52<return>" and "P", the rest is output from ODT.)

   The system will write the crashdump and then automatically restart.

5. You should now have the crash dump in [0,1]CRASH.SYS, so you can either 
extract that file and the OS image (the RSTS "SIL" file), or the whole disk.  
To analyze it, you can use the standard utility ANALYS, or the unsupported SDA 
Forth program.  I haven't tried SDA on a V6C image, though; it will probably 
work but no guarantees.  ANALYSIS needs to be the version that came with that 
particular OS release, so probably the best way to do it is to boot your system 
disk image copy on SIMH and run it there.

The key question in this instance is who called LOG$CK, the "RTS originated" 
log call.  The stack should help answer that, through tracing through it is a 
bit painful since there are no explicit stack frames so the equivalent to 
"print me the call stack" does not exist.  If the error originated in a .ERLOG 
user mode EMT, that means the RTS was unhappy.  The RTS itself isn't included 
in the crash dump, but the saved user mode PC will indicate where the error 
came from and looking at the BASIC.RTS image will indicate what it was doing at 
the time.

        paul

Reply via email to