Re: `make check' failure in 2.1.93

[SL Baur <steve at turbolinux dot co dot jp>]

[I tried building the same source on ia64 to see if this is a general
64-bit architecture problem, but could not get it to compile.]


I have a new backtrace:

(gdb) where
#0  0x200000531d8 in __kill () at __kill:2
#1  0x20000052ee8 in raise (sig=6) at ../sysdeps/posix/raise.c:27
#2  0x20000054d74 in abort () at ../sysdeps/generic/abort.c:88
#3  0x200000a4100 in __libc_malloc (bytes=2199025785400) at malloc.c:2726
#4  0x200000a8aac in tr_mallochook (size=32640, caller=0x2000003c78c)
    at mtrace.c:163
#5  0x200000a3f04 in __libc_malloc (bytes=2199025785400) at malloc.c:2691
#6  0x2000003c78c in __gconv_open (toset=0x11ffe6328 "UTF-8//", 
    fromset=0x11ffe6318 "ISO-8859-4//", handle=0x11ffe6370, flags=0)
    at gconv_open.c:263
#7  0x2000003b98c in iconv_open (tocode=0x11ffe6328 "UTF-8//", 
    fromcode=0x11ffe6318 "ISO-8859-4//") at iconv_open.c:88
#8  0x120000800 in main () at tst-loading.c:143
#9  0x2000003b470 in __libc_start_main (main=0x120000720 <main>, argc=1, 
    ubp_av=0x11fffeac8, init=0x1200005d0 <_init>, fini=0x120000a20 <_fini>, 
    rtld_fini=0x72bb, stack_end=0x11fffeab0)
    at ../sysdeps/generic/libc-start.c:111

The `bytes' parameter value looks like a valid pointer address.  I saw
similar output tracing a problem in rpm where 32 bit ints were passed
to functions expecting 64 bit long ints.

I found a warning message during compilation of malloc.c:

malloc.s: Assembler messages:
malloc.s:12129: Warning: .space repeat count is zero, ignored

This is the definition of arena_key:

.stabs "arena_key:S221",40,0,1527,arena_key
        .type    arena_key,@object
        .size    arena_key,0
        .align 2           <===== line 12129
arena_key:
        .zero   0

there is a difference between how this symbol appears in nm output
between Alpha and ia32:

Alpha:
0000000000000008 s arena_key

ia32:
00000004 b arena_key

Lastly, the problem goes away if I invoke tst-loading directly instead
of via `ld-linux.so.2 tst-loading'.  It also goes away when I try to
run the program in the debugger.