egcs sig 11 error

egcs sig 11 error

[Jon Barnett]

I am currently running Linux 2.0.33 on a pentium 120MHz system.  To date I
have had no problems with it.  I have successfully built and installed
previous versions of egcs on the system (egcs-1.0.2 and egcs-1.0.3a).  All
software (including operating system and utilities) running on the system
has been compiled with egcs-1.0.2 or 1.0.3a.

When building egcs-1.1a, I receive a sig 11.  I am using the "make
bootstrap" option.  The sig 11 always occurs during the crtbegin.o compile,
even after following the test instructions in the "SIG 11 FAQ" (so I am
assuming that it isn't a hardware problem).  I used "../configure
--prefix=/usr" in objdir to create the Makefile for the build and the
target was correctly identified as linux i586 libc1.

The tail of the results of the build are listed below:

make[1]: Entering directory `/home/users/jbarnett/egcs-1.1a/objdir/etc'
make[1]: Nothing to be done for `all'.
make[1]: Leaving directory `/home/users/jbarnett/egcs-1.1a/objdir/etc'
make[1]: Entering directory `/home/users/jbarnett/egcs-1.1a/objdir/gcc'
if [ -f libgcc2.ready ] ; then \
        true; \
else \
        touch libgcc2.ready; \
fi
/home/users/jbarnett/egcs-1.1a/objdir/gcc/xgcc
-B/home/users/jbarnett/egcs-1.1a/objdir/gcc/  -DIN_GCC   -DUSE_GNULIBC_1 -g
-O2 -I./include  -I. -I../../gcc -I../../gcc/config  -g0 \
  -finhibit-size-directive -fno-inline-functions -fno-exceptions
-fno-omit-frame-pointer \
  -c ../../gcc/crtstuff.c -DCRT_BEGIN -o crtbegin.o
xgcc: Internal compiler error: program cc1 got fatal signal 11
make[1]: *** [crtbegin.o] Error 1
make[1]: Leaving directory `/home/users/jbarnett/egcs-1.1a/objdir/gcc'
make: *** [all-gcc] Error 2


Has anyone else encountered this problem?  Is there something I have done
incorrectly?

Thanks for any help with this.

Jon Barnett

Re: egcs sig 11 error

[Jeffrey A Law]

  In message < 3.0.5.32.19980905175916.00795210@daedalus.hsys.com.au >you write:
  > I am currently running Linux 2.0.33 on a pentium 120MHz system.  To date I
  > have had no problems with it.  I have successfully built and installed
  > previous versions of egcs on the system (egcs-1.0.2 and egcs-1.0.3a).  All
  > software (including operating system and utilities) running on the system
  > has been compiled with egcs-1.0.2 or 1.0.3a.
  > 
  > When building egcs-1.1a, I receive a sig 11.  I am using the "make
  > bootstrap" option.  The sig 11 always occurs during the crtbegin.o compile,
  > even after following the test instructions in the "SIG 11 FAQ" (so I am
  > assuming that it isn't a hardware problem).  I used "../configure
  > --prefix=/usr" in objdir to create the Makefile for the build and the
  > target was correctly identified as linux i586 libc1.
Well, myself (and others) have successfully built on i586 gnulibc1.
So I'm not sure what your problem is.  

  > /home/users/jbarnett/egcs-1.1a/objdir/gcc/xgcc
  > -B/home/users/jbarnett/egcs-1.1a/objdir/gcc/  -DIN_GCC   -DUSE_GNULIBC_1 -g
  > -O2 -I./include  -I. -I../../gcc -I../../gcc/config  -g0 \
  >   -finhibit-size-directive -fno-inline-functions -fno-exceptions
  > -fno-omit-frame-pointer \
  >   -c ../../gcc/crtstuff.c -DCRT_BEGIN -o crtbegin.o
  > xgcc: Internal compiler error: program cc1 got fatal signal 11
Try running this command by hand, but add "-save-temps" to the command
line.  That will create crtstuff.i

Then

gdb cc1
r -O2 -finhibit-size-directive -fno-inline-funtions -fno-exceptions -fno-omit-frame-pointer  crtstuff.i

And see what's happening.

It's not clear if this is happening in stage1 or in a later stage of
compilation.  If it is happening in stage1, it is possible that whatever
compiler you're using to build egcs-1.1 has mis-compiled egcs-1.1.  

jeff

Re: egcs sig 11 error

[Jon Barnett]

At 00:10 8/09/98 -0600, Jeffrey A Law wrote:
>Try running this command by hand, but add "-save-temps" to the command
>line.  That will create crtstuff.i
>
>Then
>
>gdb cc1
>r -O2 -finhibit-size-directive -fno-inline-funtions -fno-exceptions
-fno-omit-frame-pointer  crtstuff.i
>
>And see what's happening.
>
>It's not clear if this is happening in stage1 or in a later stage of
>compilation.  If it is happening in stage1, it is possible that whatever
>compiler you're using to build egcs-1.1 has mis-compiled egcs-1.1.  

Thanks for the pointers.  The stage1 compiler is failing to build
crtbegin.o, it would seem.

The compiler I am using to start the bootstrap is egcs-1.0.3

$ gcc -v
Reading specs from /usr/lib/gcc-lib/i586-pc-linux-gnulibc1/egcs-2.90.29/specs
gcc version egcs-2.90.29 980515 (egcs-1.0.3 release)

Running gdb on it produces the following, with an apparent segmentation fault.

$ gdb cc1
GNU gdb 4.17
Copyright 1998 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 "i586-pc-linux-gnulibc1"...
Breakpoint 1 at 0x8048e00
Breakpoint 2 at 0x8048fd0
Breakpoint 3 at 0x80646cf: file ../../gcc/toplev.c, line 2114.
<ions -fno-exceptions -fno-omit-frame-pointer  crtstuff.i                     
Starting program: /home/users/jbarnett/egcs-1.1a/objdir/gcc/cc1 -O2
-finhibit-size-directive -fno-inline-funtions -fno-exceptions
-fno-omit-frame-pointer  crtstuff.i
Breakpoint 1 at 0x4005f778
Breakpoint 2 at 0x4005fa08
cc1: Invalid option `-fno-inline-funtions'
 __do_global_dtors_aux
Program received signal SIGSEGV, Segmentation fault.
0x8126f8a in reload (first=0x81c92e4, global=1, dumpfile=0x0)
    at ../../gcc/reload1.c:1340
1340                          in_max = MAX (in_max,
insn_needs.in_addr[j].groups[i]);

I'll have to spend some more time looking at this, so I probably won't have
a good shot at this until the weekend.

Jon.

Re: egcs sig 11 error

[Jeffrey A Law]

  In message < 3.0.5.32.19980908225932.00834ac0@daedalus.hsys.com.au >you write:
  > >It's not clear if this is happening in stage1 or in a later stage of
  > >compilation.  If it is happening in stage1, it is possible that whatever
  > >compiler you're using to build egcs-1.1 has mis-compiled egcs-1.1.  
  > 
  > Thanks for the pointers.  The stage1 compiler is failing to build
  > crtbegin.o, it would seem.
My money would be on stage1/cc1 being mis-compiled by egcs-1.0.3.

  > Program received signal SIGSEGV, Segmentation fault.
  > 0x8126f8a in reload (first=0x81c92e4, global=1, dumpfile=0x0)
  >     at ../../gcc/reload1.c:1340
  > 1340                          in_max = MAX (in_max,
  > insn_needs.in_addr[j].groups[i]);
  > 
  > I'll have to spend some more time looking at this, so I probably won't have
  > a good shot at this until the weekend.
You might want to look at the values of i & j to make sure they are
reasonable.  You may also want to look the value of reload_n_operands.

"j" should be less than or equal to reload_n_operands.  I would expect
"i" to have a value less than 15.

Assuming i & j are reasonable, look at the value of
insn_needs.in_addr[j].groups[i]

jeff

Re: egcs sig 11 error

[Jon Barnett]

At 22:29 8/09/98 -0600, Jeffrey A Law wrote:
>You might want to look at the values of i & j to make sure they are
>reasonable.  You may also want to look the value of reload_n_operands.
>
>"j" should be less than or equal to reload_n_operands.  I would expect
>"i" to have a value less than 15.
>
>Assuming i & j are reasonable, look at the value of
>insn_needs.in_addr[j].groups[i]
>

The following is the output of the debug session:

Program received signal SIGSEGV, Segmentation fault.
0x8126f8a in reload (first=0x81c92e4, global=1, dumpfile=0x0)
    at ../../gcc/reload1.c:1340
1340                          in_max = MAX (in_max,
insn_needs.in_addr[j].groups[i]);
6: &insn_needs.in_addr[j].groups[i] = (int *) 0xbfffd614
5: insn_needs.in_addr[j].groups[i] = 0
4: in_max = 0
3: reload_n_operands = 1
2: j = 0
1: i = 0
(gdb) bt
#0  0x8126f8a in reload (first=0x81c92e4, global=1, dumpfile=0x0)
    at ../../gcc/reload1.c:1340
#1  0x8119849 in global_alloc (file=0x0) at ../../gcc/global.c:587
#2  0x80669b9 in rest_of_compilation (decl=0x81ef268)
    at ../../gcc/toplev.c:3723
#3  0x805748e in finish_function (nested=0) at ../../gcc/c-decl.c:7182
#4  0x8049897 in yyparse () at c-parse.y:349
#5  0x8065522 in compile_file (name=0xbffffc96 "crtstuff.i")
    at ../../gcc/toplev.c:2768
#6  0x80680f3 in main (argc=7, argv=0xbffffb10, envp=0xbffffb30)
    at ../../gcc/toplev.c:4838
#7  0x804908e in _start ()


I had expected an out-of-bounds condition but from the debug, everything
seems ok.  I will try to roll back to an egcs-1.0.2 build and see what
happens when I build egcs-1.1a with that.

Here's hoping.

Jon.

Re: egcs sig 11 error

[Jeffrey A Law]

  In message < 3.0.5.32.19980912090300.007a2d80@daedalus.hsys.com.au >you write:
  > The following is the output of the debug session:
  > 
  > Program received signal SIGSEGV, Segmentation fault.
  > 0x8126f8a in reload (first=0x81c92e4, global=1, dumpfile=0x0)
  >     at ../../gcc/reload1.c:1340
  > 1340                          in_max = MAX (in_max,
  > insn_needs.in_addr[j].groups[i]);
  > 6: &insn_needs.in_addr[j].groups[i] = (int *) 0xbfffd614
  > 5: insn_needs.in_addr[j].groups[i] = 0
  > 4: in_max = 0
  > 3: reload_n_operands = 1
  > 2: j = 0
  > 1: i = 0
  > (gdb) bt
  > #0  0x8126f8a in reload (first=0x81c92e4, global=1, dumpfile=0x0)
  >     at ../../gcc/reload1.c:1340
  > #1  0x8119849 in global_alloc (file=0x0) at ../../gcc/global.c:587
  > #2  0x80669b9 in rest_of_compilation (decl=0x81ef268)
  >     at ../../gcc/toplev.c:3723
  > #3  0x805748e in finish_function (nested=0) at ../../gcc/c-decl.c:7182
  > #4  0x8049897 in yyparse () at c-parse.y:349
  > #5  0x8065522 in compile_file (name=0xbffffc96 "crtstuff.i")
  >     at ../../gcc/toplev.c:2768
  > #6  0x80680f3 in main (argc=7, argv=0xbffffb10, envp=0xbffffb30)
  >     at ../../gcc/toplev.c:4838
  > #7  0x804908e in _start ()
  > 
  > 
  > I had expected an out-of-bounds condition but from the debug, everything
  > seems ok.  I will try to roll back to an egcs-1.0.2 build and see what
  > happens when I build egcs-1.1a with that.
Quite odd.  Is the first stage being built with optimization?  If so
the segfault could actually be for one of the other nearby accesses.

jeff