[Bug regression/16088] New: Generates invalid code

[Bug regression/16088] New: Generates invalid code

[lucier at math dot purdue dot edu]

This one's a mystery.  Here are the versions of gcc and of gdb.  First, can I
trust what this gdb reports for this code?

[xsun21:~/programs/gambc40b4-devel] lucier% /pkgs/gcc-mainline/bin/gcc -v
Reading specs from /pkgs/gcc-mainline/lib/gcc/powerpc-apple-darwin7.4.0/3.5.0/specs
Configured with: ../configure --prefix=/pkgs/gcc-mainline
Thread model: posix
gcc version 3.5.0 20040617 (experimental)
[xsun21:~/programs/gambc40b4-devel] lucier% gdb -v
GNU gdb 5.3-20030128 (Apple version gdb-309) (Thu Dec  4 15:41:30 GMT 2003)
Copyright 2003 Free Software Foundation, Inc.
GDB is free software, covered by the GNU General Public License, and you are
welcome to change it and/or distribute copies of it under certain conditions.
Type "show copying" to see the conditions.
There is absolutely no warranty for GDB.  Type "show warranty" for details.
This GDB was configured as "powerpc-apple-darwin".

The input files are

http://www.math.purdue.edu/~lucier/GNATS/GNATS-10

The command line is

gcc -I../include -I. -no-cpp-precomp -Wall -W -Wno-unused -O1 -fno-math-errno
-fschedule-insns2 -fno-trapping-math -fno-strict-aliasing -fomit-frame-pointer
-fPIC -fno-common -g -save-temps -DHAVE_CONFIG_H -D___PRIMAL -D___LIBRARY
-D___GAMBCDIR=\"/usr/local/\" -c _nonstd.c

Here are the steps up to the invalid memory access, displaying ___fp at each step:

15546      ___SET_R4(___STRINGLENGTH(___R2))
1: ___fp = (<invalid type code 7> *) 0x60106d
(gdb) s
15547      ___IF(___NOT(___FIXLT(___FIX(0L),___R4)))
1: ___fp = (<invalid type code 7> *) 0x60106d
(gdb) s
15550      ___SET_STK(1,___FIXSUB(___R4,___FIX(1L)))
1: ___fp = (<invalid type code 7> *) 0x60106d
(gdb) s
15552      ___SET_STK(1,___BOOLEAN(___CHAREQP(___STK(-1),___STK(1))))
1: ___fp = (<invalid type code 7> *) 0x60106d
(gdb) s
15551      ___SET_STK(1,___STRINGREF(___R2,___STK(1)))
1: ___fp = (<invalid type code 7> *) 0x60106d
(gdb) s
15550      ___SET_STK(1,___FIXSUB(___R4,___FIX(1L)))
1: ___fp = (<invalid type code 7> *) 0x60106d
(gdb) s
15551      ___SET_STK(1,___STRINGREF(___R2,___STK(1)))
1: ___fp = (<invalid type code 7> *) 0x60106d
(gdb) s
15552      ___SET_STK(1,___BOOLEAN(___CHAREQP(___STK(-1),___STK(1))))
1: ___fp = (<invalid type code 7> *) 0x2f
(gdb) s
15551      ___SET_STK(1,___STRINGREF(___R2,___STK(1)))
1: ___fp = (<invalid type code 7> *) 0x2f
(gdb) s
15552      ___SET_STK(1,___BOOLEAN(___CHAREQP(___STK(-1),___STK(1))))
1: ___fp = (<invalid type code 7> *) 0xbe
(gdb) s
15557      ___IF(___FALSEP(___R1))
1: ___fp = (<invalid type code 7> *) 0xfffffffa
(gdb) s
15552      ___SET_STK(1,___BOOLEAN(___CHAREQP(___STK(-1),___STK(1))))
1: ___fp = (<invalid type code 7> *) 0xfffffffa
(gdb) 
15553      ___SET_STK(2,___R1)
1: ___fp = (<invalid type code 7> *) 0xfffffffa
(gdb) s
15555      ___SET_STK(1,___STK(2))
1: ___fp = (<invalid type code 7> *) 0xfffffffa
(gdb) 
15557      ___IF(___FALSEP(___R1))
1: ___fp = (<invalid type code 7> *) 0xfffffffa
(gdb) 
15572      ___SET_R4(___FIXSUB(___R4,___R3))
1: ___fp = (<invalid type code 7> *) 0xfffffffa
(gdb) 

Program received signal EXC_BAD_ACCESS, Could not access memory.
0x0009cc2c in ___H__20___nonstd (___ps=0x1e31f8) at _nonstd.c:15572
15572      ___SET_R4(___FIXSUB(___R4,___R3))
1: ___fp = (<invalid type code 7> *) 0x28

The assembly is

	.stabd	68,0,15546
	lwz r0,-1(r25)
	.stabd	68,0,15547
	rlwinm. r10,r0,25,7,29
	ble- cr0,L4572
	.stabd	68,0,15550
	addi r11,r10,-4
	.stabd	68,0,15552
	lwz r0,4(r30)
	.stabd	68,0,15551
	srawi r9,r11,1
	.stabd	68,0,15550
	stw r11,-4(r30)
	.stabd	68,0,15551
	add r9,r9,r25
	lhz r2,3(r9)
	.stabd	68,0,15552
	mr r9,r30
	.stabd	68,0,15551
	slwi r2,r2,2
	addi r2,r2,2
	stw r2,-4(r30)
	.stabd	68,0,15552
	xor r2,r2,r0
	addic r2,r2,-1
	subfe r2,r2,r2
	rlwinm r2,r2,0,0,29
	addi r2,r2,-2
	.stabd	68,0,15557
	cmpwi cr7,r2,-2
	.stabd	68,0,15552
	stwu r2,-4(r9)
	.stabd	68,0,15553
	stw r11,-8(r30)
	.stabd	68,0,15555
	stw r11,0(r9)
	.stabd	68,0,15557
	bne+ cr7,L4578
	b L4577
L4572:
	.stabd	68,0,15562
	stwu r28,-4(r30)
	.stabd	68,0,15564
	mr r11,r10
	mr r9,r30
	b L4578
L4577:
	.stabd	68,0,15568
	mr r11,r10
L4578:
	.stabd	68,0,15572
	lwz r2,0(r9)

The invalid access is occuring at the source line

   ___r4=((((int)(((((unsigned int)((*((int*)((___fp[-(0)])-(1))))))>>(3
+5))>>1)))<<2));

which is why I traced the value of ___fp.

It seems that ___fp has a rational value until line 15552, then all hell breaks
loose (if I can believe gdb).

And I haven't been able to find a smaller example, sorry.

This works with all previous versions of gcc.

Brad

-- 
           Summary: Generates invalid code
           Product: gcc
           Version: 3.5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: regression
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: lucier at math dot purdue dot edu
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: powerpc-apple-darwin7.4.0
  GCC host triplet: powerpc-apple-darwin7.4.0
GCC target triplet: powerpc-apple-darwin7.4.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16088

[Bug rtl-optimization/16088] Generates invalid code

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|regression                  |rtl-optimization
           Keywords|                            |wrong-code


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16088

[Bug rtl-optimization/16088] [3.5 Regression] Generates invalid code

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-08-03 23:39 -------
Note the debuging problem for "invalid type" is filed under PR 14860.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|Generates invalid code      |[3.5 Regression] Generates
                   |                            |invalid code
   Target Milestone|---                         |3.5.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16088

[Bug rtl-optimization/16088] [4.0 Regression] Generates invalid code

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-09-15 03:59 -------
Does this work now?

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16088

[Bug rtl-optimization/16088] [4.0 Regression] Generates invalid code

[lucier at math dot purdue dot edu]

------- Additional Comments From lucier at math dot purdue dot edu  2004-09-15 14:00 -------
Subject: Re:  [4.0 Regression] Generates invalid code

> Does this work now?

Does it fail at the same spot?  No.

Does it fail in a similar way at a different spot?  Yes.

This is with

[descartes:~/programs/gambc40b4-devel] lucier% gcc -v
Reading specs from 
/pkgs/gcc-mainline/lib/gcc/powerpc-apple-darwin7.5.0/4.0.0/specs
Configured with: ../configure --prefix=/pkgs/gcc-mainline 
--with-gmp=/pkgs/gmp-4.1.3 --with-mpfr=/pkgs/gmp-4.1.3
Thread model: posix
gcc version 4.0.0 20040913 (experimental)

Should I file a new PR?

And, can I trust what information gdb gives me when tracing codes 
compiled with this version of gcc on powerpc-apple-darwin7.5.0?

Brad



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16088

[Bug rtl-optimization/16088] [4.0 Regression] Generates wrong code

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[4.0 Regression] Generates  |[4.0 Regression] Generates
                   |invalid code                |wrong code


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16088

[Bug rtl-optimization/16088] [4.0 Regression] Generates wrong code

[geoffk at gcc dot gnu dot org]

------- Additional Comments From geoffk at gcc dot gnu dot org  2004-10-27 23:53 -------
No, I don't think you can trust the values of variables in gdb when using stabs (like on Darwin) at 
optimisation levels higher than -O0.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16088

[Bug rtl-optimization/16088] [4.0 Regression] Generates wrong code

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-10-29 13:19 -------
Can you try again, there was recently patch which might help, also can you attach a preprocessed 
source which you can link?

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16088

[Bug rtl-optimization/16088] [4.0 Regression] Generates wrong code

[lucier at math dot purdue dot edu]

------- Additional Comments From lucier at math dot purdue dot edu  2004-11-04 03:31 -------
Subject: Re:  [4.0 Regression] Generates wrong code

I tried again, I'm having the same problems with today's 4.0.

I also have the same problems on sparc-sun-solaris2.9.  I tried 
debugging it with gdb 6.2.1 on that platform but got garbage 
information about variables, etc.  Geoff says I can't trust gdb on 
powerpc-apple-darwin7.5.0, either, with -O1.

Is there any gdb on any OS platform that I can use to debug this 
problem?  Should I try it on a friend's x86 Linux box?  Or should I 
proceed in a different way.



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16088

[Bug rtl-optimization/16088] [4.0 Regression] Generates wrong code

[lucier at math dot purdue dot edu]

------- Additional Comments From lucier at math dot purdue dot edu  2004-11-18 01:13 -------
Maybe this bug should be closed for lack of specificity.

The problem is the following: gcc-4.0 miscompiles C code generated by the
Scheme->C compiler Gambit-C (http://www.iro.umontreal.ca/~gambit).  I have not
been able to debug the problem (all the various version of gdb I tried have not
been helpful).  In the meantime, two code generation bugs have been fixed (one
in the past few days), and several still remain.

So the specific information (such as it is) in this bug report is no longer
accurate.  It is still a fact that gcc-4.0 miscompiles code generated by
Gambit-C (and examples of this are easy to find, e.g., the compiler itself
aborts in many cases)---perhaps this should be a placeholder bug report for this
general problem?  I don't know.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16088

[Bug rtl-optimization/16088] [4.0 Regression] Generates wrong code

[giovannibajo at libero dot it]

------- Additional Comments From giovannibajo at libero dot it  2004-11-18 01:26 -------
OK closing this as invalid then. It would be great if you could reduce it to 
something attachable. A big preprocessed source code which segfaults is already 
a starting point, for instance.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |INVALID


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16088