NetBSD Problem Report #8952

Received: (qmail 5445 invoked from network); 4 Dec 1999 19:41:59 -0000
Message-Id: <199912041941.OAA205647@antharia.mit.edu>
Date: Sat, 4 Dec 1999 14:41:56 -0500 (EST)
From: Dan Winship <danw@MIT.EDU>
Reply-To: danw@MIT.EDU
To: gnats-bugs@gnats.netbsd.org
Subject: gcc on powerpc sometimes generates bad struct copies
X-Send-Pr-Version: 3.95

>Number:         8952
>Category:       port-powerpc
>Synopsis:       gcc on powerpc sometimes generates bad struct copies
>Confidential:   no
>Severity:       serious
>Priority:       low
>Responsible:    port-powerpc-maintainer
>State:          closed
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Sat Dec 04 11:42:01 +0000 1999
>Closed-Date:    Mon Nov 19 21:53:18 +0000 2001
>Last-Modified:  Mon Nov 19 21:53:18 +0000 2001
>Originator:     Dan Winship
>Release:        current
>Organization:
>Environment:
System: NetBSD frotz 1.4M NetBSD 1.4M (FROTZ) #19: Fri Dec 3 14:22:05 EST 1999 root@frotz:/usr/src/sys/arch/macppc/compile/FROTZ macppc


>Description:
	gcc on the powerpc outputs floating-point instructions to do
	64-bit copies, but doesn't check that the data is properly
	aligned.
>How-To-Repeat:

danw@frotz:/tmp> cat foo.c
int main(int argc, char **argv)
{
  struct foo { short a, b, c, d; } one;
  struct bar { short e; struct foo f; } two;

  two.f = one;
}
danw@frotz:/tmp> gcc -o foo foo.c
danw@frotz:/tmp> ./foo
Bus error (core dumped)

>Fix:
	Fix gcc to know it can't do that. (I think it actually only
	needs to be 32-bit aligned: Changing the shorts to ints makes
	it work.)
>Release-Note:
>Audit-Trail:
State-Changed-From-To: open->closed 
State-Changed-By: matt 
State-Changed-When: Mon Nov 19 13:47:03 PST 2001 
State-Changed-Why:  
Bug no longer present in gcc version egcs-2.91.66 19990314 (egcs-1.1.2 release) 
>Unformatted:

Home	PR Database Search