#328 gdb access violation when debugging

[Vladimir Michl]

Hello dovelopers,

I compile some C++ program which have too much classes
(such of this have all members static) with mingw 2.0.0
(gcc 3.2)
mingw-runtime 2.4, w32api 2.2.

Flags when compiling is
-mms-bitfields -mwin32 -mno-cygwin -Wall $(VERSION) -g
-mconsole

linked libraries are
-lkernel32 -ladvapi32 -lws2_32 -lstdc++

When I try trace program in main() (with next command)
on second line gdb cause access violation.

Command 'dissasemble main' cause access vialation too.

I think that gdb has corrupted symbol table, because
access violation is caused when show (in disassemble)
call to ___chkstk function (show to objdump -d cipsrvr.exe)

I attach compiled binary which cause gdb access violation.

I try gdb 5.1.1 and gdb 5.2.1 with the same result.

[Danny Smith]

Logged In: YES
user_id=11494

There is an _alloca/_chkstck bug in gcc3.2 for cygwin
and mingw that is exposed at -O2 and higher when
large arrays or structs (> 4kb) are allocated on the stack
This _may_ be the problem. That bug is fixed now.

Danny

[Vladimir Michl]

Logged In: YES
user_id=213474

Now is in gcc 3.2.2-20030208 downloadable from sourceforge?

I try compile binary without -O2 only with -g.

The result is the same -> access violation.

I try gcc 3.2 and gcc 3.2.2-20030208.

If binary is compiled without -g and without -O2 or with
-O2, main may be disassembled without crash. (gcc 3.2.2).

Because sourceforge.net accept attached files only to 256KB
size, compiled zipped binary which crash gdb may be
downloaded from
http://hlubocky.del.cz/~michlv/mingw/CIPSRVR.ZIP.

[Danny Smith]

Logged In: YES
user_id=11494

Sorry, I wasn't clear. The problem may be that gdb was
compiled with 3.2.0 or 3.1 so has stack corruption fault.
Danny

[Vladimir Michl]

Logged In: YES
user_id=213474

gcc 3.2.2

my program compiled with gcc 3.2.2.
gdb 5.2.1 compiled with gcc 3.2.2. configured with:
./configure under msys-1.0.8.

after run gdb on gdb with my compiled program and type:

in debugged gdb:
(gdb) disassemble main
Dump of assembler code for function main:
0x401636 <main>: push %ebp
0x401637 <main+1>: mov %esp,%ebp
0x401639 <main+3>: push %edi
0x40163a <main+4>: push %esi
0x40163b <main+5>: push %ebx
0x40163c <main+6>: sub $0x9c,%esp
0x401642 <main+12>: and $0xfffffff0,%esp
0x401645 <main+15>: mov $0x0,%eax
0x40164a <main+20>: mov %eax,0xffffff80(%ebp)
0x40164d <main+23>: mov 0xffffff80(%ebp),%eax
0x401650 <main+26>: call 0x426650
During symbol reading, inner block (0x40dc92-0x40df2e) not
inside outer block (0
x0-0x21).

In top-gdb:
Program received signal SIGSEGV, Segmentation fault.
finish_block (symbol=0x0, listhead=0x5aed90,
old_blocks=0xabababab, start=0,
end=33, objfile=0x10b3d88) at buildsym.c:388
388 if (BLOCK_SUPERBLOCK (pblock->block) ==
NULL)

(top-gdb) bt
#0 finish_block (symbol=0x0, listhead=0x5aed90,
old_blocks=0xabababab,
start=0, end=33, objfile=0x10b3d88) at buildsym.c:388
#1 0x004b0525 in process_one_symbol (type=36, desc=0,
valu=33,
name=0x1190028 "", section_offsets=0x10b8f70,
objfile=0x10b3d88)
at dbxread.c:2758
#2 0x004afc43 in read_ofile_symtab (pst=0x13b79c0) at
dbxread.c:2602
#3 0x004af77f in dbx_psymtab_to_symtab_1
(pst=0x13b79c0) at dbxread.c:2430
#4 0x004af715 in dbx_psymtab_to_symtab_1
(pst=0x13b9278) at dbxread.c:2416
#5 0x004af715 in dbx_psymtab_to_symtab_1
(pst=0x138b960) at dbxread.c:2416
#6 0x004af893 in dbx_psymtab_to_symtab (pst=0x138b960)
at dbxread.c:2471
#7 0x00405b13 in psymtab_to_symtab (pst=0x138b960) at
symfile.c:374
#8 0x004678e8 in blockvector_for_pc_sect (pc=4351568,
section=0x0,
pindex=0x22f7a4, symtab=0x0) at blockframe.c:585
#9 0x00467948 in block_for_pc_sect (pc=4351568,
section=0x0)
at blockframe.c:644
#10 0x004679a3 in find_pc_sect_function (pc=4351568,
section=0x0)
at blockframe.c:665
#11 0x0044d972 in build_address_symbolic (addr=4351568,
do_demangle=0,
name=0x22f868, offset=0x22f858, filename=0x22f864,
line=0x22f85c,
unmapped=0x22f860) at printcmd.c:647
#12 0x0044d84e in print_address_symbolic (addr=4351568,
stream=0x3d3d70,
do_demangle=0, leadin=0x44dc46 " ") at printcmd.c:559
#13 0x0044dc77 in print_address (addr=4351568,
stream=0x3d3d70)

[Wayne M]

Logged In: YES
user_id=1313445
Originator: NO

reporter states "I do not use any of that any more. Possibly I do not even have the original sources"