bootstrap/8146: gcc 3.2 miscompiles gcc 2.95.3

bootstrap/8146: gcc 3.2 miscompiles gcc 2.95.3

[prj]

>Number:         8146
>Category:       bootstrap
>Synopsis:       gcc 3.2 miscompiles gcc 2.95.3
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          wrong-code
>Submitter-Id:   net
>Arrival-Date:   Fri Oct 04 10:46:00 PDT 2002
>Closed-Date:
>Last-Modified:
>Originator:     Paul Jarc
>Release:        3.2
>Organization:
>Environment:
System: Linux multivac.cwru.edu 2.4.19 #1 SMP Mon Aug 5 11:15:13 EDT 2002 i686 unknown
Architecture: i686

	
host: i686-pc-linux-gnu
build: i686-pc-linux-gnu
target: i686-pc-linux-gnu
configured with: ../gcc-3.2/configure --prefix=/package/host/localhost/gcc-3.2 --with-as=/package/host/localhost/gcc-3.2/conf/as --with-ld=/package/host/localhost/gcc-3.2/conf/ld --enable-threads --enable-libgcj --disable-nls
glibc 2.2.5
binutils 2.13
>Description:
	If I start with gcc 3.2 and try to bootstrap gcc 2.95.3, I get
	a bootstrap comparison failure.  The documentation says that this
	usually indicates that the host compiler (gcc 3.2 here) miscompiled
	gcc (2.95.3).
	Bootstrap comparison failure!
	cccp.o differs
	f/data.o differs
	f/target.o differs
	make[1]: *** [compare-lean] Error 1
	3.1 and 3.1.1 also fail to bootstrap 2.95.3.  IIRC, 3.1 had
	differences only in cccp.o.
	I suggest adding bootstrapping old versions to the pre-release
	testing procedure.  Having a buggy compiler is bad; having one
	that is buggy in such a way as to complicate downgrading to a
	not-known-to-me-to-be-buggy version is worse.
	(I was able to downgrade by doing "make install" after the failing
	"make bootstrap", and then using that incomplete 2.95.3 to do a
	successful bootstrap of a new 2.95.3.)
>How-To-Repeat:
	Build 2.95.3 using 3.2.  I think you already have the relevant
	source files. :)  Files from my build or installation
	directories of 3.2 or 2.95.3 are available on request.
>Fix:
	
>Release-Note:
>Audit-Trail:
>Unformatted:

Re: bootstrap/8146: gcc 3.2 miscompiles gcc 2.95.3

[Richard Henderson]

The following reply was made to PR bootstrap/8146; it has been noted by GNATS.

From: Richard Henderson <rth@redhat.com>
To: Joe Buck <jbuck@synopsys.com>
Cc: gcc-gnats@gcc.gnu.org, gcc-bugs@gcc.gnu.org, hjl@gnu.org, prj@po.cwru.edu
Subject: Re: bootstrap/8146: gcc 3.2 miscompiles gcc 2.95.3
Date: Wed, 11 Dec 2002 15:36:00 -0800

 On Fri, Nov 08, 2002 at 05:04:38PM -0800, Joe Buck wrote:
 > This PR is now fixed on the branch, but not in the mainline; I have just 
 > verified
 > that we still get the same failure.
 > 
 > Should HJ's patch be applied to the mainline as well?
 
 No.  A different patch was intended to fix this problem on
 mainline.
 
 
 r~

Re: bootstrap/8146: gcc 3.2 miscompiles gcc 2.95.3

[Joe Buck]

The following reply was made to PR bootstrap/8146; it has been noted by GNATS.

From: Joe Buck <jbuck@synopsys.com>
To: gcc-gnats@gcc.gnu.org, gcc-bugs@gcc.gnu.org, hjl@gnu.org, prj@po.cwru.edu
Cc:  
Subject: Re: bootstrap/8146: gcc 3.2 miscompiles gcc 2.95.3
Date: Fri, 08 Nov 2002 17:04:38 -0800

 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=8146
 
 This PR is now fixed on the branch, but not in the mainline; I have just 
 verified
 that we still get the same failure.
 
 Should HJ's patch be applied to the mainline as well?
 
 I will be taking this PR as [mainline regression] now.
 

Re: bootstrap/8146: gcc 3.2 miscompiles gcc 2.95.3

[Joe Buck]

The following reply was made to PR bootstrap/8146; it has been noted by GNATS.

From: Joe Buck <jbuck@synopsys.com>
To: gcc-gnats@gcc.gnu.org, gcc-bugs@gcc.gnu.org, prj@po.cwru.edu
Cc:  
Subject: Re: bootstrap/8146: gcc 3.2 miscompiles gcc 2.95.3
Date: Fri, 08 Nov 2002 13:54:38 -0800

 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=8146
 
 This is fixed on the branch by HJ's patch:
 
 2002-11-07  H.J. Lu <hjl@gnu.org>
 
          Fix bootstrap/8146 and
          gcc.c-torture/execute/20020307-2.c with -march=i686.
 
          * calls.c (PUSH_ARGS_REVERSED): Define only if not defined.
          * expr.c (PUSH_ARGS_REVERSED): Likewise.
 
          * config/i386/i386.h (PUSH_ARGS_REVERSED): Set to 1.
 
 Does this bug affect the mainline?
 

Re: bootstrap/8146: gcc 3.2 miscompiles gcc 2.95.3

[Joe Buck]

The following reply was made to PR bootstrap/8146; it has been noted by GNATS.

From: Joe Buck <jbuck@synopsys.com>
To: gcc-gnats@gcc.gnu.org, gcc-bugs@gcc.gnu.org, prj@po.cwru.edu
Cc:  
Subject: Re: bootstrap/8146: gcc 3.2 miscompiles gcc 2.95.3
Date: Thu, 31 Oct 2002 14:13:52 -0800

 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=8146
 
 I'm marking this as high priority, even though it's not certain that 
 this is a
 3.2 bug; it could be a 2.95.3 bug.  At least, it should be investigated 
 in more
 detail.
 

Re: bootstrap/8146: gcc 3.2 miscompiles gcc 2.95.3

[Janis Johnson]

The following reply was made to PR bootstrap/8146; it has been noted by GNATS.

From: Janis Johnson <janis187@us.ibm.com>
To: gcc-gnats@gcc.gnu.org, gcc-prs@gcc.gnu.org, gcc-bugs@gcc.gnu.org,
   nobody@gcc.gnu.org, prj@po.cwru.edu
Cc:  
Subject: Re: bootstrap/8146: gcc 3.2 miscompiles gcc 2.95.3
Date: Wed, 23 Oct 2002 16:38:23 -0700

 I tried this (bootstrapping GCC 2.95.3) with 3.2 that I built from a
 release tarball and with the 3.2 branch updated an hour or so ago, and
 with both of them I get the same bootstrap comparison failures that
 were reported in the PR:
 
   Bootstrap comparison failure!
   cccp.o differs
   f/data.o differs
   f/target.o differs
 
 This was on i686-pc-linux-gnu, with an old Caldera distribution plus
 binutils 2.12.1.
 
 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=8146