strange number corruption on pdp11/34
bqt at update.uu.se
Thu Jan 29 18:13:44 CST 2015
On 2015-01-29 19:32, Jacob Ritorto wrote:
> Noel wrote: I'd really recommend fixing at least one more board set -
> otherwise, when the
> next failure happens, you'll have _no_ working board sets...
> Yep, you're right. *sigh* more work. But that's honestly a good chunk
> of the fun of this hobby, right? I guess I should take advantage of the
> momentum and curiosity about this bug that we have going here. I just need
> quite a lot of hand holding because I'm still way too green with this stuff.
So all other combinations results in different types of strange
behavior. So you might as well start listing what happens with all the
> Johnny, you're insisting that I put in the real numbers for address and csr
> for testing the drive (for instance). I'm going to do that next, here.
> But are you understanding that some of us think that the reason it prompts
> a zero default is that it's a manifestation of the zero bug and that the
> real value *is* actually safe but hidden in memory? Did you see the
> RSX-11M crash dump I posted in the other thread?
I have not yet even understood what the "zero" bug means. :-)
Are you saying that the system works fine, but for some magic reason all
numbers printed out comes out as zero, no matter what number you tried
What I do remember is that all XXDP tests that I can ever recall running
actually shows, and needs proper CSRs and vectors. And when they show 0,
they use 0. But who knows, I might be remembering wrong, or that
specific test might be the exception. Or any number of possible
explanations. I did check the documentation for the tests, and they also
suggest that the test should show the actual CSR and vector, so the fact
that you get zeroes there are either some really weird magic, or else
the tests are clever and do not give your standard controller numbers
because XXDP booted from that device, and it tries to avoid you
accidentally destroying your diagnostics pack...
> And to your other query, yes, I'm booting from the rl02 I'm testing with.
> I just switch out to a scratch pack once the diag is loaded and ready to
> run (hence the warnings about writing on LoaD device).
And I wonder if the test might try to be clever and avoid just
defaulting to writing on the diagnostics pack...
By the way. I don't remember if I saw exactly which test your were
running that failed.
> On Thu, Jan 29, 2015 at 8:48 AM, Johnny Billquist <bqt at update.uu.se> wrote:
>> Short question: What did you boot from? The same RL controller and disk
>> you then wanted to run diagnostics on? Or did you have a second storage,
>> from which you booted?
>> On 2015-01-29 04:29, Jacob Ritorto wrote:
>>> From what I observed, this zeroed-out prompt was just another
>>> of the zeroes bug. The values were there in memory and were supposed to
>>> offered to the user but got blanked out with zeroes instead. I had
>>> suspected this and just hit enter at the prompt and XXDP did, in fact,
>>> exercise the disk. There were blinking lights and head noise, obvious
>>> being done that wouldn't have, had there actually been real zeroes in the
>>> address and csr fields. I might swap the bad cpu board back in and try
>>> again, but now that the problem has been isolated and removed, I'm
>>> to just enjoy the pdp11 :)
>>> On Wed, Jan 28, 2015 at 10:19 PM, Rick Murphy <rick at rickmurphy.net>
>>> At 07:34 PM 1/27/2015, Jacob Ritorto wrote:
>>>> Johnny, I just took the defaults,
>>>> BUS ADDRESS (O) 0 ?
>>>> What Johnny said. The prompt says "(O)", not "(0)" - the letter O for
>>>> Octal. The zero after that isn't a default, it simply means that the
>>>> hasn't been set.
>>>> There is no default. Or at least no sane one. Zero in this case means
>>>> it's writing to memory address zero for CSR. The diagnostic should reject
>>>> that, but diagnostics are notoriously user unfriendly.
Johnny Billquist || "I'm on a bus
|| on a psychedelic trip
email: bqt at softjar.se || Reading murder books
pdp is alive! || tryin' to stay hip" - B. Idol
More information about the cctalk