Enter your search terms Submit search form
	Web pdp-10.trailing-edge.com

22-Apr-88 23:36:27-PDT,2065;000000000011
Return-Path: <[email protected]>
Received: from XX.LCS.MIT.EDU by SRI-NIC.ARPA with TCP; Fri 22 Apr 88 23:36:23-PDT
Date: Sat, 23 Apr 1988  02:38 EDT
Message-ID: <[email protected]>
From: Rob Austein <[email protected]>
To:   [email protected]
cc:   [email protected]
Subject: I know you said no more KCC for a while, but...

Here's some more thoughts on extended addressing, PSECTs, etc.
They make use of a LINK switch I never thought I'd have a good use
for.  The following builds a reasonable looking HELLO.EXE from a
HELLO.REL produced with KCC-5 right out of the box:

; [XX.LCS.MIT.EDU]XX:<SRA>HELLO.CCL.1, 23-Apr-88 00:57:50, Edit by SRA
; -*- CMD -*- mode for EMACS.
; This is a test file to see how loading HELLO.C works with
; various PSECT and extended addressing frobs.
;
; Here's where CC.EXE invents interesting parameters (Z$foo should not
; be confused with the E$foo symbols in CENDPS.C)
/define:b$data:#1001000/define:z$data:#1377777
/define:b$code:#1400000/define:z$code:#1777777
;
; Boiler plate
/set:data:b$data/limit:data:z$data
/set:code:b$code/limit:code:z$code
/redirect:data:code/symseg:psect:data
;
; Files to read and write.
hello/save
c:libckx,hello
;
; Show relocations counters (after hello.rel, before libc.rel) and finish up
/count/go

Features of this approach:

- You could dispose of LIBCKX.REL, do /DEFINE:$$SECT:1, etc, instead.
- You can get rid of the hardwired assumption that the code PSECT
  starts at location 400000 (gag me with a...).
- You could even define the location of things like the PDL this way,
  on the fly (well, in section zero you couldn't but that's ok, we
  already know the right way to define PDL and malloc space in section
  zero).

I was pleasantly surprised to find that (1) the CRT code correctly
handled the case where it discovered that it was already in section
one, and (2) that the END <n,,entvec> in the main module actually did
the right thing (LINK is cleverly fixing up the entry vector address
as a 30 bit value).
23-Apr-88 00:21:40-PDT,868;000000000001
Return-Path: <[email protected]>
Received: from XX.LCS.MIT.EDU by SRI-NIC.ARPA with TCP; Sat 23 Apr 88 00:21:38-PDT
Date: Sat 23 Apr 88 03:24:20-EDT
From: Rob Austein <[email protected]>
Subject: Addendum
To: [email protected]
cc: [email protected]
Message-ID: <[email protected]>

One more thought and I have to sign off for the night.  If you are interested
in the idea of using /DEFINE to flush LIBCKX, here's a technique
that might be useful.

Have KCC emit, to LINK:

/DEFINE:$$$KLX:n

where n is zero or one (might not want to use $$SECT for this), then
in the file that defines all the byte pointers and such, do

$BPxxxx==:<sect_zero_value+<<extended_value - sect_zero_value>*$$$KLX>

and let LINK sort it all out as a Polish expression.  Voila, no more
need for LIBCKX.  Ain't arithmetic wonderful.

--Rob
-------