Under IRIX (6.5.20) Python 2.3 (final) crashes with 
a core dump while executing a SCons job 
(www.scons.org). I've recompiled Python with -g but 
the behavior is the same. Here is the beginning of 
the dbx stack trace:

dbx version 7.3 MR 55458_Apr30_MR Apr 30 1999 
13:44:41
dbx Warning: Unknown processor type 0xe, 
assuming this is not an R8000
Core from signal SIGSEGV: Segmentation violation
(dbx) where
>  0 mark_save(state = <no such address>, lo = 
<no such address>, hi = 0, mark_stack_base = 
(nil)) ["/txusr/rwgrosse/selfx/Python-
Modules/_sre.c":289, 0x10107bc4]
   1 sre_match(state = 0x10840c78, pattern = 
0x10207ef0, level = 669) 
["/txusr/rwgrosse/selfx/Python-
Modules/_sre.c":979, 0x10109d74]
   2 sre_match(state = 0x10840c78, pattern = 
0x10207f10, level = 668) 
["/txusr/rwgrosse/selfx/Python-
Modules/_sre.c":1263, 0x1010ad24]
   3 sre_match(state = 0x10840c78, pattern = 
0x10207ef0, level = 667) 
["/txusr/rwgrosse/selfx/Python-
Modules/_sre.c":991, 0x10109ea0]
   4 sre_match(state = 0x10840c78, pattern = 
0x10207f10, level = 666) 
["/txusr/rwgrosse/selfx/Python-
Modules/_sre.c":1263, 0x1010ad24]
   5 sre_match(state = 0x10840c78, pattern = 
0x10207ef0, level = 665) 
["/txusr/rwgrosse/selfx/Python-
Modules/_sre.c":991, 0x10109ea0]
   6 sre_match(state = 0x10840c78, pattern = 
0x10207f10, level = 664) 
["/txusr/rwgrosse/selfx/Python-
Modules/_sre.c":1263, 0x1010ad24]

The alternation of _sre.c:991, _sre.c:1263 continues 
to the end of the dbx output which stops at 99; the 
value of level is decreasing monotonously.

To reproduce upload this file to an IRIX machine:

http://cci.lbl.gov/cctbx_build/send_octet_stream.cgi?
source=results/2003_08_04_2056/cctbx_python_23_
bundle.selfx

In a new, empty directory enter the command:

perl cctbx_python_23_bundle.selfx

When prompted enter the number of available CPU's 
and wait for the Python 2.3 installation to finish 
(log is in the file py_install_log). The newly 
compiled Python is used to start a SCons build 
procedure. The crash occurs a few files into the 
process of compiling C++ sources.

I've tried the exact same source code under Redhat 
8.0. There it finishes without a crash. Therefore I 
am thinking you will need an IRIX machine to 
reproduce the crash.