6.0 BUG? SIGSEGV under cygwin

6.0 BUG? SIGSEGV under cygwin

[Pierre Muller]

  I compiled the 6.0 branch GDB
with -gstabs+  debug option.

  I think that there is a problem 
in the current main and 6.0 branches,
regarding use of sym_private file of struct objfile.

 This field is used to stored two incompatible 
things :
  in dwarf2-frame.c it is assumed that 
this field contains a struct dwarf2_fde pointer,
but the same field is also used in 
coffread.c and xcoffread.c
to store a struct coff_symfile_info pointer.

  These two structures are completely incompatible and lead to the crash 
with the following backtrace :
(gdb) bt 6
#0  dwarf2_frame_find_fde (pc=0x10ad7068) at ../../src/gdb/dwarf2-frame.c:1038
#1  0x005542e2 in dwarf2_frame_cache (next_frame=0x10121ba0,
    this_cache=0x10121c0c) at ../../src/gdb/dwarf2-frame.c:505
#2  0x0055469e in dwarf2_frame_this_id (next_frame=0x10121ba0,
    this_cache=0x10121c0c, this_id=0x10121c28)
    at ../../src/gdb/dwarf2-frame.c:600
#3  0x0046d980 in get_frame_id (fi=0x10121bf8) at ../../src/gdb/frame.c:243
#4  0x004706fa in get_prev_frame (this_frame=0x10121bf8)
    at ../../src/gdb/frame.c:1924
#5  0x0042963c in backtrace_command_1 (count_exp=0x0, show_locals=0,
    from_tty=1) at ../../src/gdb/stack.c:1215
(More stack frames follow...)

the source listing looks like this :
(gdb) li -20
1018       inital location associated with it into *PC.  */
1019
1020    static struct dwarf2_fde *
1021    dwarf2_frame_find_fde (CORE_ADDR *pc)
1022    {
1023      struct objfile *objfile;
1024
1025      ALL_OBJFILES (objfile)
1026        {
1027          struct dwarf2_fde *fde;
(gdb)
1028          CORE_ADDR offset;
1029
1030          offset = ANOFFSET (objfile->section_offsets, SECT_OFF_TEXT (objfile));
1031
1032          fde = objfile->sym_private;
1033          while (fde)
1034            {
1035              if (*pc >= fde->initial_location + offset
1036                  && *pc < fde->initial_location + offset + fde->address_range)

(gdb) p  objfile.name
$30 = 0x10a44430 "/cygdrive/f/pas/cvs100/bin/CYGWIN1.DLL"

p *fde makes no sense,
but 
p *(struct coff_symfile_info *)fde
does make sense:
(gdb) p *fde
$28 = {cie = 0x0, initial_location = 0, address_range = 1627394048,
  instructions = 0xc9018 <Address 0xc9018 out of bounds>, end = 0x0,
  next = 0x0}
(gdb) p /x *(struct coff_symfile_info *)fde
$29 = {min_lineno_offset = 0x0, max_lineno_offset = 0x0,
  textaddr = 0x61001000, textsize = 0xc9018, stabsects = 0x0,
  stabstrsect = 0x0, stabstrdata = 0x0}
Which looks rather reasonable to people who know
cygwin a little : 0x61000000 is the default base address of the cygwin DLL.

I have really no idea how to fix the problem,
but I hope that my message is clear enough.

The following steps should be enough to reproduce the problem,
on a cygwin system only, of course.

::::1) recompile the gdb with -gstabs+ debug option.
::::2) run the created gdb on itself.
./gdb ./gdb in build/gdb dir.
::::::3) set a breakpoint in command_loop
(top-gdb) b command_loop
::::::4) run the inferior
(top-gdb) run

(top-gdb) run
Starting program: /home/muller/gdb/gdb60/build/gdb/gdb.exe

warning: obsolete '/home/muller/gdb.ini' found. Rename to '/home/muller/.gdbinit
'.
GNU gdb 5.3.90_2003-07-15-cvs adapted to Free Pascal
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 "i686-pc-cygwin".
Breakpoint 3, command_loop () at ../../src/gdb/top.c:760
760       int stdin_is_tty = ISATTY (stdin);
::::::5) ask for a backtrace
(top-gdb) bt
#0  command_loop () at ../../src/gdb/top.c:760
Segmentation fault (core dumped)

I used an older gdb to be able to find out the problem.....

I hope this will be fixed before 6.0 release....





Pierre Muller
Institut Charles Sadron
6,rue Boussingault
F 67083 STRASBOURG CEDEX (France)
mailto:muller@ics.u-strasbg.fr
Phone : (33)-3-88-41-40-07  Fax : (33)-3-88-41-40-99

Re: 6.0 BUG? SIGSEGV under cygwin

[Daniel Jacobowitz]

On Tue, Jul 15, 2003 at 04:40:56PM +0200, Pierre Muller wrote:
>   I compiled the 6.0 branch GDB
> with -gstabs+  debug option.
> 
>   I think that there is a problem 
> in the current main and 6.0 branches,
> regarding use of sym_private file of struct objfile.
> 
>  This field is used to stored two incompatible 
> things :
>   in dwarf2-frame.c it is assumed that 
> this field contains a struct dwarf2_fde pointer,
> but the same field is also used in 
> coffread.c and xcoffread.c
> to store a struct coff_symfile_info pointer.

Mark Kettenis posted a patch for this just a few days ago on
gdb-patches.  Would you test it?  See:
  Subject: [RFC/RFA] Per-objfile data mechanism
From Sunday.


-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: 6.0 BUG? SIGSEGV under cygwin

[Pierre Muller]

At 16:51 15/07/2003, Daniel Jacobowitz wrote:
>On Tue, Jul 15, 2003 at 04:40:56PM +0200, Pierre Muller wrote:
>>   I compiled the 6.0 branch GDB
>> with -gstabs+  debug option.
>> 
>>   I think that there is a problem 
>> in the current main and 6.0 branches,
>> regarding use of sym_private file of struct objfile.
>> 
>>  This field is used to stored two incompatible 
>> things :
>>   in dwarf2-frame.c it is assumed that 
>> this field contains a struct dwarf2_fde pointer,
>> but the same field is also used in 
>> coffread.c and xcoffread.c
>> to store a struct coff_symfile_info pointer.
>
>Mark Kettenis posted a patch for this just a few days ago on
>gdb-patches.  Would you test it?  See:
>  Subject: [RFC/RFA] Per-objfile data mechanism

I just tested this patch and it does indeed solve the problem.
Thanks Mark!
This (or something equivalent) should really be 
incorporated into 6.0 branch...


Pierre Muller
Institut Charles Sadron
6,rue Boussingault
F 67083 STRASBOURG CEDEX (France)
mailto:muller@ics.u-strasbg.fr
Phone : (33)-3-88-41-40-07  Fax : (33)-3-88-41-40-99