better backtrace for emacs

[Marcus Brinkmann]

Hi,

here is a much better backtrace, with debugging symbols.

More importantly, I have went through both glibc in parallel,
in two debugging sessions, and found the place where it differs.

In the working version:

0x0816db9f      70          cp = (char *) (*real_morecore) (0);
1: x/i $eip  0x816db9f <check_memory_limits+55>:        call   *%eax
(gdb)
__default_morecore (increment=422313983) at ../sysdeps/generic/morecore.c:46
46      ../sysdeps/generic/morecore.c: No such file or directory.
        in ../sysdeps/generic/morecore.c
1: x/i $eip  0x1377d14 <__default_morecore>:    push   %ebp
(gdb)
0x01377d15      46      in ../sysdeps/generic/morecore.c
1: x/i $eip  0x1377d15 <__default_morecore+1>:  mov    %esp,%ebp

etc.

But in the non-working version:
0x0816db9f      70          cp = (char *) (*real_morecore) (0);
2: x/i $eip  0x816db9f <check_memory_limits+55>:        call   *%eax
(gdb)
0x01377d14 in free_atfork (mem=0x192bffff, caller=0x83e8ec0) at malloc.c:4885
4885    malloc.c: No such file or directory.
        in malloc.c
2: x/i $eip  0x1377d14 <free_atfork+168>:
    je     0x1377d30 <free_atfork+196>

etc.  So, obviously, the "real_morecore" variable contains a bogus value
in emacs.  It turns out that the value seems to be hard coded in the
binary.  Setting real_morecore to __default_morecore in
realloc_glyph_pool made it work!
__morecore had the correct value of __default_morecore.

I set a break point at r_alloc_reinit but it was not called before
it segfaults.  So maybe it is not called early enough?  Mmh, this malloc
mangling is still pretty obscure to me.  Do I have enough info for the
emacs people (or do you know a fix)?

#0  0x012f7ba8 in strfnames () from /root/debug/libc.so.0.3
#1  0x01377d2d in free_atfork (mem=0x192bffff, caller=0x83e8ec0)
    at ../mach/lock-intern.h:86
#2  0x013752e6 in chunk_alloc (ar_ptr=0x1442540, nb=40008) at malloc.c:2593
#3  0x01374c7a in __libc_malloc (bytes=40000) at malloc.c:2810
#4  0x0811a60c in emacs_blocked_malloc (size=40000) at alloc.c:737
#5  0x01374bed in __libc_malloc (bytes=40000) at malloc.c:2797
#6  0x0811a27f in xmalloc (size=40000) at alloc.c:520
#7  0x080508ad in realloc_glyph_pool (pool=0x83a6b40, matrix_dim=
    {width = 80, height = 25}) at dispnew.c:1628
#8  0x080511ac in adjust_frame_glyphs_for_frame_redisplay
    (f=0x8290a40) at dispnew.c:1628
#9  0x08050e66 in adjust_frame_glyphs (f=0x8290a40) at dispnew.c:2141
#10 0x08050e30 in adjust_frame_glyphs_initially () at dispnew.c:2127
#11 0x08057742 in init_display () at dispnew.c:6537
#12 0x080d5ed5 in main (argc=3, argv=0x1023c00, envp=0x1023c10) at emacs.c:1448
#13 0x0131ddb6 in __libc_start_main (main=0x80d5744 <main>, argc=3,
    ubp_av=0x1023c00, init=0x804d6d0 <_init>, fini=0x81739f8 <_fini>,
    rtld_fini=0xb440 <_dl_fini>, stack_end=0x1023bfc)
    at ../sysdeps/generic/libc-start.c:129

eax            0x2a410  173072
ecx            0xcccccccd       -858993459
edx            0x0      0
ebx            0xccccccc        214748364
esp            0x1023938        0x1023938
ebp            0x1023978        0x1023978
esi            0x2a000  172032
edi            0x8417000        138506240
eip            0x12f7ba8        0x12f7ba8
eflags         0x10202  66050
cs             0x17     23
ss             0x1f     31
ds             0x1f     31
es             0x1f     31
fs             0x1f     31
gs             0x1f     31
fctrl          0x0      0
fstat          0x0      0
ftag           0x0      0
fiseg          0x0      0
fioff          0x0      0
foseg          0x0      0
fooff          0x0      0
fop            0x0      0

(gdb) x/20i $pc
0x12f7ba8 <strfnames+113512>:   jmp    *0x464(%ebx)
0x12f7bae <strfnames+113518>:   push   $0x8b0
0x12f7bb3 <strfnames+113523>:   jmp    0x12f6a38 <strfnames+109048>
0x12f7bb8 <strfnames+113528>:   jmp    *0x468(%ebx)
0x12f7bbe <strfnames+113534>:   push   $0x8b8
0x12f7bc3 <strfnames+113539>:   jmp    0x12f6a38 <strfnames+109048>
0x12f7bc8 <strfnames+113544>:   jmp    *0x46c(%ebx)

etc