Bug 28183

Summary:	[4.0/4.1/4.2/4.3 regression] assembler error "FATAL: can't close x.o" on m68k with new binutils
Product:	gcc	Reporter:	Martin Michlmayr <tbm>
Component:	target	Assignee:	Not yet assigned to anyone <unassigned>
Status:	RESOLVED INVALID
Severity:	normal	CC:	debian-gcc, gcc-bugs, schwab, wouter
Priority:	P5	Keywords:	wrong-code
Version:	4.1.1
Target Milestone:	4.0.4
Host:		Target:	m68k-linux-gnu
Build:	m68k-linux-gnu	Known to work:	3.4.6
Known to fail:	4.0.4 4.1.1 4.2.0	Last reconfirmed:
Attachments:	preprocessed source

Description Martin Michlmayr 2006-06-27 21:17:14 UTC

[ forwarded from http://bugs.debian.org/375519 ]

We get the following assembler error when compiling the application cln.  The error goes away when you drop -fPIC or when you use gcc 3.4.  Additionally, the error does not show up with older releases of binutils.  While 2.16.91 20060413 produces the error, 2.16.91 20051117 works fine.

I'm not sure if this is a bug in binutils but given that gcc 3.4 works I'm filing it as a gcc bug.  Andreas, since you're also involved in binutils, I hope you know where to put the blame. ;-)

crest% g++-4.1 -fPIC -DPIC -c cl_MI.ii
crest% g++-4.1 -fPIC -DPIC -c -O2 cl_MI.ii
cl_MI.o: No error
/tmp/ccjCp4mS.s: Assembler messages:
/tmp/ccjCp4mS.s:18716: FATAL: can't close cl_MI.o
: No error
crest% g++-4.0 -fPIC -DPIC -c -O2 cl_MI.ii
cl_MI.o: No error
/tmp/ccreQBwI.s: Assembler messages:
/tmp/ccreQBwI.s:19590: FATAL: can't close cl_MI.o
: No error
crest% g++-3.4 -fPIC -DPIC -c -O2 cl_MI.ii
crest% as --version
GNU assembler 2.16.91 20060413 Debian GNU/Linux
Copyright 2005 Free Software Foundation, Inc.

Comment 1 Martin Michlmayr 2006-06-27 22:08:58 UTC

Created attachment 11766 [details]
preprocessed source

Comment 2 Martin Michlmayr 2006-06-27 22:21:22 UTC

What I said about binutils might be wrong.  It seems it's not due to a difference in the version, but due to one being native and one being cross.  I just upgraded the binutils-m68k on i386 to the same version as that on m68k native and strangely enough the same .s file works on i386 with m68k-as but fails on m68k:

native:
crest% as test.s
a.out: No error
test.s: Assembler messages:
test.s:18960: FATAL: can't close a.out
: No error
crest% as --version
GNU assembler 2.16.91 20060413 Debian GNU/Linux
Copyright 2005 Free Software Foundation, Inc.
This program is free software; you may redistribute it under the terms of
the GNU General Public License.  This program has absolutely no warranty.
This assembler was configured for a target of `m68k-linux-gnu'.
crest%

i386:
1173:tbm@reyes: ~] /usr/local/bin/m68k-linux-gnu-as test.s
1174:tbm@reyes: ~] /usr/local/bin/m68k-linux-gnu-as --version
GNU assembler 2.16.91 20060413 Debian GNU/Linux
Copyright 2005 Free Software Foundation, Inc.
This program is free software; you may redistribute it under the terms of
the GNU General Public License.  This program has absolutely no warranty.
This assembler was configured for a target of `m68k-linux-gnu'.
1175:tbm@reyes: ~]

Comment 3 Andrew Pinski 2006-06-28 11:15:41 UTC

(In reply to comment #2)
That would almost mean as is being miscompiled on m68k.

Comment 4 Andreas Schwab 2006-08-07 23:12:40 UTC

The removed comment says:

-  /* If will do cse, generate all results into pseudo registers
-     since 1) that allows cse to find more things
-     and 2) otherwise cse could produce an insn the machine
-     cannot support.  An exception is a CONSTRUCTOR into a multi-word
-     MEM: that's much more likely to be most efficient into the MEM.
-     Another is a CALL_EXPR which must return in memory.  */

So it looks like point 2 is true.

Comment 5 Andreas Schwab 2006-08-07 23:13:15 UTC

Sorry, wrong bug.

Comment 6 Roman Zippel 2007-01-31 18:20:34 UTC

This bug can be closed, it's really a bug in binutils triggered by some inline assembly in cln. Here is a bit more info:
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=388000

Comment 7 Andrew Pinski 2007-01-31 18:27:25 UTC

Marking as invalid as requested.