Test cases: dining philosophers, race condition and deadlock?

[Nikolaus Waxweiler]

Hi all,
can somebody confirm that the dining philosophers at  
http://svn.gna.org/viewcvs/*checkout*/gnustep/libs/base/trunk/Testing/diningPhilosophers.m?rev=22472  
deadlock after some time?

I use gnustep-base-1.13.0 on FreeBSD 6.1, GCC 3.4.4 and compiled the file  
with the following GNUmakefile:

-------------------------------------
include $(GNUSTEP_MAKEFILES)/common.make

#
# Tool
#
VERSION = 0.1
PACKAGE_NAME = ToolTest
TOOL_NAME = ToolTest
ToolTest_TOOL_ICON =
GNUSTEP_INSTALLATION_DIR = $(HOME)/GNUstep/Tools

#
# Libraries
#
ToolTest_TOOL_LIBS +=

#
# Class files
#
ToolTest_OBJC_FILES = \
diningPhilosophers.m

-include GNUmakefile.preamble
include $(GNUSTEP_MAKEFILES)/aggregate.make
include $(GNUSTEP_MAKEFILES)/tool.make
-include GNUmakefile.postamble
------------------------------------

It doesn't always deadlock immediately, I got the following after some  
waiting:

-----------
Philosopher 3 is eating.
Philosopher 3 is eating.
Philosopher 3 is eating.
Philosopher 3 is eating.
Philosopher 3 is eating.
Philosopher 3 is eating.
Philosopher 0 can start eating.
Philosopher 0 is eating.
<long pause..>
Philosopher 3 is eating.
<another long pause...>
^C
------------

By the way, what use is a test that doesn't end :O? Oh, and: the other  
testfiles (e.g. the one for arrays) look like they need a real unit  
testing library. Or is there some "Error:" grepping involved?