Re: Crash when launched -nw on Mac OS X, compiled with gcc 4.0.1

[Peter Dyballa]

Am 24.07.2006 um 16:42 schrieb Richard Stallman:

>     When launched in Terminal GNU Emacsen 22.0.50 and 23.0.0 both  
> crash
>     after some time. In gdb GNU Emacs continues until:
>
>             Program received signal EXC_BAD_INSTRUCTION, Illegal instruction/
>     operand.
>             0x001f26d8 in Fsqrt (arg=501192) at floatfns.c:561
>             561       IN_FLOAT (d = sqrt (d), "sqrt", arg);
>
> Please examine relevant variables (you can start with d)
> to determine what's actually happening in the function.
>
>     After a few 'make distclean ; ./configure ... ; make' cycles it  
> seems
>     to be this gcc option that makes GNU Emacs crash when running  
> with no
>     windows: -mpowerpc-gpopt. This is again true for both GNU Emacsen
>     22.0.50 and 23.0.0. More optimisation like -O2 works.
>
> This could be a GCC bug.  However, the only way to track it down
> is to find the place where the compiled code is wrong.  To do that,
> you need to debug the program in the usual way.
>
> If you put a breakpoint at Fsqrt and step thru, what do you see?

        pete 274 /\ gdb emacs
	GNU gdb 6.3.50-20050815 (Apple version gdb-477) (Sun Apr 30 20:06:22  
GMT 2006)
        Copyright 2004 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"...Reading symbols  
for shared libraries ................. done
        
        DISPLAY = :0.0
        TERM = xterm-color
        Breakpoint 1 at 0x1f27d4: file floatfns.c, line 556.
        Breakpoint 2 at 0x151c2c: file sysdep.c, line 1393.
        (gdb) run
        Starting program: /Users/pete/Quellen/Emacs_CVS/emacs/src/emacs -Q -nw
        Reading symbols for shared libraries ..++++++ done
        No struct type named Lisp_Symbol.
        (gdb) cont
        Continuing.
        
        
        
        Breakpoint 1, Fsqrt (arg=4216608) at floatfns.c:556
        556       double d = extract_float (arg);
        (gdb) step
        extract_float (num=2) at floatfns.c:219
        219     {
        (gdb) step
        220       CHECK_NUMBER_OR_FLOAT (num);
        (gdb) step
        222       if (FLOATP (num))
        (gdb) step
        224       return (double) XINT (num);
        (gdb) step
        225     }
        (gdb) step
        Fsqrt (arg=4216608) at floatfns.c:561
        561       IN_FLOAT (d = sqrt (d), "sqrt", arg);
        (gdb) step
        
	Program received signal EXC_BAD_INSTRUCTION, Illegal instruction/ 
operand.
        0x001f2828 in Fsqrt (arg=4216608) at floatfns.c:561
        561       IN_FLOAT (d = sqrt (d), "sqrt", arg);
        (gdb) quit
        
        The program is running.  Exit anyway? (y or n) y

When hitting the breakpoint d is 7.1699012343114344e-308 (__DBL_MIN__  
is #define's as 2.2250738585072014e-308), when I hit the bad  
statement d has become 527076.

Can I do some more?


The error might come from my side! Reading once more the Apple docs  
on gcc 4 it seems that my PowerPC processor, PPC 7447A, is a 64 bit  
model and by using -mpowerpc I put into 32 bit mode. And this  
contradicts the -mcpu=7450 setting (7450 is the correct cpu model for  
the 7447A). I made again distclean and re-configured again -- and  
again the same error:

        Breakpoint 1, Fsqrt (arg=4216608) at floatfns.c:556
        556       double d = extract_float (arg);
        (gdb) step
        extract_float (num=2) at floatfns.c:219
        219     {
        (gdb) step
        220       CHECK_NUMBER_OR_FLOAT (num);
        (gdb) step
        222       if (FLOATP (num))
        (gdb) step
        224       return (double) XINT (num);
        (gdb) step
        225     }
        (gdb) step
        Fsqrt (arg=4216608) at floatfns.c:561
        561       IN_FLOAT (d = sqrt (d), "sqrt", arg);
        (gdb) step
        
	Program received signal EXC_BAD_INSTRUCTION, Illegal instruction/ 
operand.
        0x001f2828 in Fsqrt (arg=4216608) at floatfns.c:561
        561       IN_FLOAT (d = sqrt (d), "sqrt", arg);

d is the same as above when reaching the breakpoint and when the bad  
instruction strikes.

--
Greetings

  Pete

The future will be much better tomorrow.
                               -- George W. Bush