Strange LibreOffice crash on i386

Hi,

The libreoffice package in pkgsrc-wip builds fine on -current/amd64 but one command
fails with a signal 11 error on i386:

[...]
/usr/pkgsrc/wip/libreoffice/work/libreoffice-bootstrap-2011-06-15/solver/350/unxnbsd.pro/bin/uno -ro uno_services.rdb -ro uno_types.rdb -s com.sun.star.test.bridge.BridgeTest -- com.sun.star.test.bridge.CppTestObject
[1]    954 segmentation fault (core dumped)   -ro uno_services.rdb -ro uno_types.rdb -s com.sun.star.test.bridge.BridgeTes

I tried to debug it in gdb, and got a completely different error: a breakpoint
in the dynamic loader:

(gdb) run
Starting program:
/usr/pkgsrc/wip/libreoffice/work/libreoffice-bootstrap-2011-06-15/solver/350/unxnbsd.pro/bin/uno
-ro uno_services.rdb -ro uno_types.rdb -s com.sun.star.test.bridge.BridgeTest
-- com.sun.star.test.bridge.CppTestObject
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)

Program received signal SIGTRAP, Trace/breakpoint trap.
0xbbbf9a77 in _sys___sigprocmask14 () from /usr/libexec/ld.elf_so
(gdb) bt
#0  0xbbbf9a77 in _sys___sigprocmask14 () from /usr/libexec/ld.elf_so
#1  0xbbbf2817 in _rtld_call_init_functions () from /usr/libexec/ld.elf_so
#2  0xbb894400 in ?? ()
(gdb) 

Now, a breakpoint in the i386 ld.so binary seems very dubious to me. Can gdb
be really trusted on -current ?

If yes, I have no clue what is going on here. All ideas welcome.

-- 
Francois Tigeot