Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site arizona.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!mhuxr!ihnp4!arizona!budd
From: budd@arizona.UUCP (tim budd)
Newsgroups: net.lang.st80
Subject: I am not an identifier
Message-ID: <22796@arizona.UUCP>
Date: Sat, 16-Mar-85 14:27:56 EST
Article-I.D.: arizona.22796
Posted: Sat Mar 16 14:27:56 1985
Date-Received: Sun, 17-Mar-85 02:42:52 EST
Distribution: net
Organization: Dept of CS, U of Arizona, Tucson
Lines: 24

<who are you?  I am the new number two>

Here is a mystery I would eagerly like an answer to.

The Little Smalltalk system has a feature whereby the current environment
can be saved in a file and later restored.  It does this by basically
copying everything from the start of the users data space (found by
carefully inserting a global declaration in the first location of the first
loaded module) to _end.  This works fine on the 780 *.  On several other
machines, however, (examples are 11/70, DecPro 350, IBM PC/IX, PE 11/32),
the next time one does a ``system()'' call following the load one receives
the cryptic message
	xxx: is not an identifier
from the sh fired off by the system, and the call doesn't work.  Even
stranger, on at least one machine (the 11/70) this does not happen if
the program is given by a path, only if it is in the current directory.

Obviously, something is being clobbered that shouldn't be, but what? and
where?

* note: this works because nothing on the stack points into memory at the
point the save and/or load is performed, and there are no open files.
Also clearly there are machines (such as the HP 9000) where nothing of the
sort will work.