Return-Path: daemon@media.mit.edu
Received: by media.mit.edu (5.57/DA1.0.4.amt)
	id AA15410; Sun, 11 Jun 95 09:00:38 -0400
Received: from aleve.media.mit.edu by morphine.media.mit.edu; (5.65/1.1/06Jun95-8.2MPM)
	id AA15401; Sun, 11 Jun 1995 09:00:36 -0400
Received: from media-lab.media.mit.edu by aleve.media.mit.edu; (5.65/1.1/06Jun95-8.2MPM)
	id AA01509; Sun, 11 Jun 1995 09:00:36 -0400
Received: by media-lab.media.mit.edu (5.57/DA.WS.1.0.5)
	id AA03992; Sun, 11 Jun 95 09:00:29 -0400
Message-Id: <9506111300.AA03992@media-lab.media.mit.edu>
To: rsargent@ai.media.mit.edu, anarch@ai.media.mit.edu
Cc: fredm@media.mit.edu
Subject: BUG IN PCODE SOURCE
Date: Sun, 11 Jun 95 09:00:28 -0400
From: "Fred G. Martin" <fredm@media.mit.edu>
X-Mts: smtp

Hi guys,

Wanted to let you know about a small but possibly pesky bug in the
pcode source, at least as it is distributed on cher.

The problem is the use of a "FDB" rather than a "RMB 2" when declaring
the zero-page variables.  This causes two bytes to be assembled to
location 001E, which, when downloaded, clobber the bootstrap
routine.

I discovered the bug when trying to download pcode using DLM -- it got
clobbered.  Randy, I don't know if your DL has protection against
this, or if you just got lucky and the bytes being downloaded matched
the bytes in the downloader.  Probably more likely that the bytes
downloaded to a part of code no longer in use?

Anyhow, the offending line is:

  scheduler_iterations FDB 1  ; 1e  count of # times through scheduler loop


	-Fred