[Python-bugs-list] [ python-Bugs-493815 ] IRIX 6.5 bus error

[noreply@sourceforge.net]

Bugs item #493815, was opened at 2001-12-15 21:56
You can respond by visiting: 
http://sourceforge.net/tracker/?func=detail&atid=105470&aid=493815&group_id=5470

Category: Python Interpreter Core
Group: Platform-specific
>Status: Closed
>Resolution: Duplicate
Priority: 5
Submitted By: Ralf W. Grosse-Kunstleve (rwgk)
Assigned to: Tim Peters (tim_one)
Summary: IRIX 6.5 bus error

Initial Comment:
A simple statement causes a bus error under IRIX 6.5:

% ./configure -without-threads
% make
% ./python
Python 2.1.1 (#1, Dec 15 2001, 21:40:34) [C] on irix6
Type "copyright", "credits" or "license" for more 
information.
>>> print 3. * 24
Bus error

OS version: IRIX 6.5.10m
MIPSpro Compilers: Version 7.3.1.2m

The threaded version of Python does not exhibit the 
same problem.


----------------------------------------------------------------------

>Comment By: Tim Peters (tim_one)
Date: 2001-12-16 00:12

Message:
Logged In: YES 
user_id=31435

Well, there's always *some* optimization problem with SGI's 
compiler du jour.  That's why the primary Python README 
file has said this for years:

"""
WARNING: There are bugs in the optimizer of some versions of
SGI's compilers that can cause bus errors or other strange
behavior, especially on numerical operations.  To avoid 
this, try building with "make OPT=".
"""

"Turn off optimization" is the first response to every SGI 
bug report, and is also usually the last.

I don't know whether SGI users report these bugs to SGI 
(they're not due to non-standard C in Python -- every one 
to date has been pinned on the compiler), but things never 
really improve there -- the bugs just shift around.  I wish 
SGI would add Python to their compiler regression test 
suite.

If you want to submit a patch to spray configure warnings, 
someone will look at it, but since it's already covered in 
README and the bugs differ across compiler versions, I'm 
not keen on the idea.

----------------------------------------------------------------------

Comment By: Ralf W. Grosse-Kunstleve (rwgk)
Date: 2001-12-15 23:39

Message:
Logged In: YES 
user_id=71407

Thanks for the fast response.
Compiling floatobject.c with -O0 does the trick.
It would be advantageous if configure could print
a warning about SGI problems with floatobject.c
(and didn't Python 1.5.2 have a problem with
complexobject.c?).
Ralf


----------------------------------------------------------------------

Comment By: Tim Peters (tim_one)
Date: 2001-12-15 22:04

Message:
Logged In: YES 
user_id=31435

Recompile floatobject.c with optimization disabled; the 
problem will almost certainly go away; it's an SGI compiler 
bug if so.

----------------------------------------------------------------------

You can respond by visiting: 
http://sourceforge.net/tracker/?func=detail&atid=105470&aid=493815&group_id=5470