From mboxrd@z Thu Jan  1 00:00:00 1970
From: wilson@gcc.gnu.org
To: gcc-bugs@gcc.gnu.org, gcc-prs@gcc.gnu.org, jreiser@BitWagon.com, jsm28@cam.ac.uk, nobody@gcc.gnu.org, wilson@gcc.gnu.org
Subject: Re: c/2047: Internal compiler error in `require_complete_type', at c-typeck.c:96
Date: Mon, 10 Sep 2001 18:39:00 -0000
Message-id: <20010911013944.20535.qmail@sourceware.cygnus.com>
X-SW-Source: 2001-09/msg00225.html
List-Id: <gcc-prs.sourceware.org>

Synopsis: Internal compiler error in `require_complete_type', at c-typeck.c:96

Responsible-Changed-From-To: unassigned->wilson
Responsible-Changed-By: wilson
Responsible-Changed-When: Mon Sep 10 18:39:44 2001
Responsible-Changed-Why:
    IA-64 maintainer
State-Changed-From-To: open->closed
State-Changed-By: wilson
State-Changed-When: Mon Sep 10 18:39:44 2001
State-Changed-Why:
    I reproduced the problem with the old obsolete 000717
    release.  It is a problem with cascading errors, resulting
    in a type tree that is invalid.  The internal consistency
    checking code detects the problem and triggers an abort.
    
    With current gcc sources, when we detect an inconsistency,
    if there were prior errors in the source code, then instead
    of aborting, we emit an error message "confused by earlier
    errors" and exit cleanly.  This seems like a reasonable
    solution to me.  We could add more code to handle this
    case, but it doesn't seem worthwhile, since we would just
    get even more useless error messages than we already do.

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&pr=2047&database=gcc