Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: notesfiles
Path: utzoo!watmath!clyde!cbosgd!hplabs!hp-pcd!orstcs!nathan
From: nathan@orstcs.UUCP (nathan)
Newsgroups: net.lang.c++
Subject: statics in state tables
Message-ID: <34200006@orstcs.UUCP>
Date: Wed, 12-Mar-86 00:50:00 EST
Article-I.D.: orstcs.34200006
Posted: Wed Mar 12 00:50:00 1986
Date-Received: Fri, 14-Mar-86 07:04:20 EST
Organization: Oregon State University - Corvallis, OR
Lines: 26
Nf-ID: #N:orstcs:34200006:000:1131
Nf-From: orstcs!nathan    Mar 11 21:50:00 1986

Forward-declaring static variables:

The following has been a problem in C; I wonder if it is still a
problem in C++.  Suppose you want to initialize a state table with
circular references, and want the table entries declared static.
For example:

	struct entry { entry* next; int i; };

	static entry e1 = { &e2, 0 };
	static entry e2 = { &e1, 1 };

The first initialization will fail, as e2 hasn't been seen
yet.  We need to "forward" declare e2 without defining it.  But how?
Putting an "extern" on would prevent definition, but would conflict
with the "static" later on (at least according to the ANSI C draft).
Calling them static in the forward-declaration would define them too.
Do we need a special-case here (e.g. static overrides earlier extern)?
If so, the manual should say so -- otherwise implementors will
make their own merry mess of it.

By the way, it seems to me a significant difference from C that the
default storage class for globals is "static" -- worth noting at least
in Section r.15, "Differences from C".  I only noticed it by accident.

Nathan C. Myers		hplabs!hp-pcd!orstcs!nathan  nathan@oregon-state