[COFF] Building OS from source in the olden days

[dave at horsfall.org (Dave Horsfall)]

On Wed, 11 Nov 2020, Greg 'groggy' Lehey wrote:

> I'm currently reviewing a paper about Unix and Linux, and I made the 
> comment that in the olden days the normal way to build an OS image for a 
> big computer was from source.  Now I've been asked for a reference, and 
> I can't find one!  Can anybody help?

Depends what you mean by "olden days" and "big computer".  As I recall we 
(Uni of NSW) had the source to the 360/50 and the Cyber 72, but not for 
the VMS stuff; binaries were patched with IEBUPDTE and later on SUPERZAP 
(possibly written locally).

I got an official pat on the back for getting SPITBOL to work after its 
time-bombs (yes, plural) expired[*]...

And we had the source to something called Unix Edition 5 & 6 etc, but they 
were hardly mainframes :-)

[*]
The first bomb failed with an error message, so I patched that.  It then 
started crashing rather mysteriously, and I discovered that it was taking 
an indirect jump to whatever was in R0 at the time (I think).  Rather than 
waste time digging them all out, I wrote a program that LOADed the binary, 
scanned memory for a word that matched that date, and printed each address 
so they could then be inspected by hand.  There were something like six of 
them...  One big SUPERZAP later, and we had a working SPITBOL compiler 
again; a bored CompSci student is terrible to behold.

-- Dave