[Coco] 'arc' copy command and Stack Overflow (ERROR #207)

[Gene Heskett]

On Monday 26 January 2015 09:43:15 L. Curtis Boyle did opine
And Gene did reply:
> ATD would be Allen Dekok, one of the original 4 programmers for
> NitrOS-9. So that source (unless somebody merged comments from various
> releases) would be older than the one you worked on, I believe.
> 
> L. Curtis Boyle
> curtisboyle at sasktel.net

The merge of comments I can understand, but there are several places where 
I put in 6309 code as it was faster, that instead of conditionalizing it 
for the build in question, were simply reverted so it would work on a 
6809. It works, but we need to save every cycle we can.

Moving that code to a 32 or 64 bit platform is very helpful as the coco's 
max buffer allocation of 56K simply is not enough to allow room for 
explanatory comments in any great detail.  As a consequence, comments I 
put into the original dis as I figured out what it was doing, have been 
lost, and now that I have access to an editor than can handle a 100 
megabyte file, I will have to reinvent.

These reversions can fully explain why, when I had it optimized for the 
6309, I could get consistent megaread times of 11 seconds, but the last 
time I checked, a couple years ago it was a leisurely 17 or 18 seconds!  I 
was puzzled and disappointed at the time, but did not drag out a copy of 
geany to investigate.  I should have.

Thats a huge speed hit IMNSHO.

What we need now is a code tracer that keeps a running image of the stack 
as it exists after every instruction so that an errant overwrite of an 
important registers contents can be quickly spotted.  Likewise for dead 
code in the form of a label that is never called. I am sure there is some 
of that, but identifying it is a problem on a scale that is not be 
implementable on the coco.

I have some other projects here that I need to complete (a Green & Green 
style blanket chest that I have $600+ in the wood and hinge hardware now, 
and a large check I was sent in 2009 that never arrived) before I'll get 
time to "get my head" back into this.  Whats obvious is that I sure need 
to. What bothers me is that the initials ATD are in a comment line 
immediately above some of those reversions, so they have to be newer than 
when I worked last on it at ed 34. At one point, my commented out code was 
said to be slower!  'scuse me?  It's also discouraging to have to redo 
work I spent several months doing and testing for correctness about 20 
years ago.  Done I might add, with no credits in the files header.  The 
history seems to start at ed 35, when Boisy got bit by a feature I put 
into 34 as an aid to backar, so only files modified since the last backup 
would be saved in later runs.  I asked at the time that everyone raise the 
descriptors it.sas from $08 to $10 or more, I've been using $20, sometimes 
even $FF when copying large files as it reduces fragmentation by a huge 
factor.  No one listened, including Boisy. I still think its a great idea 
for people who like to keep incremental backups as they would be 5% of the 
size of a full backup.  And I may have a way figured out to do it so the 
deaf among us will not get bit.

Cheers, Gene Heskett
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Genes Web page <http://geneslinuxbox.net:6309/gene>
US V Castleman, SCOTUS, Mar 2014 is grounds for Impeaching SCOTUS