10 May
2024
10 May
'24
11:14 p.m.
In the early '80's, I did some programming with Micro Concurrent Pascal,
on embedded CDP1802 systems. It was really nice to be able to program in
something other than assembly language (a cross-assembler that ran on a
PDP-11 system).
Regarding protections, it didn't have many. I remember spending a day
tracking down a fatal bug with a logic analyzer (emulators were still a
dream in this small company)... another programmer had used an array
subscript out of range and the compiler didn't catch it for some reason.
So in this array defined [0..20], when the typo caused a write to
FOO[60] instead of FOO[20], bad things happened.
Ah, the good old days ;)
-Charles
10 May
10 May
11:37 p.m.
New subject: Random items on Pascal #3
On Fri, 10 May 2024, Charles via cctalk wrote:
Regarding protections, it didn't have many. I
remember spending a day
tracking down a fatal bug with a logic analyzer (emulators were still a dream
in this small company)... another programmer had used an array subscript out
of range and the compiler didn't catch it for some reason. So in this array
defined [0..20], when the typo caused a write to FOO[60] instead of FOO[20],
bad things happened.
Ah, the good old days ;)
At Goddard Space Flight Center, my position was negligible (gopher and APL
and FORTRAN programming for a British pysicist studying the Van Allen
belts).
I was told that some of the many locally applied patches were done by
writes to array elements with negative subscripts.
We may have been the first one to get some IBM 360 operating systems. I
remember one time, shortly after "upgrading", we rolled back to the
previous one, until the next one arrived.
--
Grumpy Ol' Fred cisin(a)xenosoft.com
11 May
11 May
12:01 a.m.
New subject: Random items on Pascal #3
On 5/10/24 16:37, Fred Cisin via cctalk wrote:
I was told that some of the many locally applied
patches were done by
writes to array elements with negative subscripts.
CDC 6000 (the one with PPUs) OS (SCOPE, KRNONOS, MACE and NOS) used a
single PPU that, among other things, monitored the contents of location
1 in each control point's field length. Normally, it was zero; but if
it became nonzero, it took the format of a system request, very often to
be serviced by a certain PPU program.
In CDC FTN (and probably RUN), you could get the (60 bit word) address
of a variable using either the LOCF() function call or .LOC. unary
operator, depending on the dialect of FORTRAN.
So, to put something into that location 1 address word was pretty
straightforward.
INTEGER AX(1), IX;
IX = LOCF( AX)
AX ( -IX-1) = <system request.
You'd add a little loop waiting for the system request location to go
back to zero, but that was it. A lot of "quick and dirty" code was
written that way and cleaned up later for distribution.
We did the something similar with BASIC on the F-85 micro to write
directly to the CRT buffer--display something, search memory for it,
change the display to confirm and you have your video buffer address.
--Chuck
6:09 p.m.
New subject: Better demagnetize all of your cables!
https://www.swpc.noaa.gov/products/aurora-30-minute-forecast
stormy sunny weather
196
days inactive
197
days old
3 comments
3 participants
participants (3)
-
Charles -
Chuck Guzis -
Fred Cisin