Bug 481

Summary:	Broken module "multiprocessing.synchronize"
Product:	Mageia	Reporter:	Panos Christeas <xrg>
Component:	RPM Packages	Assignee:	Michael Scherer <misc>
Status:	RESOLVED FIXED	QA Contact:
Severity:	normal
Priority:	Normal	CC:	marja11
Version:	Cauldron
Target Milestone:	---
Hardware:	x86_64
OS:	Linux
Whiteboard:
Source RPM:	python-2.7.1-1.mga1.src.rpm	CVE:
Status comment:
Attachments:	This forces the semaphore functions to be available regardless of host The corresponding patch to the spec file

Description Panos Christeas 2011-03-20 21:38:18 CET

To reproduce:

python
 > from multiprocessing import synchronize
 ImportError: This platform lacks a functioning sem_open implementation, therefore, the required synchronization primitives needed will not function, see issue 3770.

According to the issue, the binary of /usr/lib64/python2.7/lib-dynload/_multiprocessing.so lacks support for semaphores (?!) . Perhaps a packaging error?

Reproducible: 

Steps to Reproduce:

Comment 1 Panos Christeas 2011-03-20 21:38:41 CET

http://bugs.python.org/issue3770

Comment 2 Michael Scherer 2011-03-21 12:21:39 CET

I think someone filed a similar bug report on mandriva. The problem is our buildbot is minimal, and do not have the proper fs mounted.

Comment 3 Panos Christeas 2011-04-07 00:38:28 CEST

Created attachment 207 [details]
This forces the semaphore functions to be available regardless of host

Comment 4 Panos Christeas 2011-04-07 00:39:07 CEST

Created attachment 208 [details]
The corresponding patch to the spec file

Comment 5 Michael Scherer 2011-04-14 16:44:09 CEST

The patch look good, expect this cannot be sent to upstream I guess. I will apply it.

Comment 6 Michael Scherer 2011-04-14 18:18:15 CEST

Ok after discussing with Victor Stinner ( python developper ), he told me the patch was not really the proper one, but as I do not have more time to fix it properly, I will take this for now ( I just hope to not forget to fix this correctly one day ).

Comment 7 Michael Scherer 2011-04-14 23:19:17 CEST

Ok, it seems now to block on test when compiling , due to the patch ( as gdb have show on the BS ).

0x00007ff4519854b0 in sem_timedwait () from /lib64/libpthread.so.0

It is blocked in threading test, on x86_64. Maybe related ( found by pterjan : http://bugs.python.org/issue5114 ).

Comment 8 Michael Scherer 2011-04-15 11:45:42 CEST

Ok so this was already fixed on mandriva : https://qa.mandriva.com/show_bug.cgi?id=62281

Comment 9 Michael Scherer 2011-05-02 16:16:27 CEST

IIRC, this is fixed, no ?

Comment 10 Marja Van Waes 2011-10-07 22:18:11 CEST

(In reply to comment #9)
> IIRC, this is fixed, no ?

@ Panos Christeas, can you please confirm or disconfirm?

CC: (none) => marja11

Comment 11 Panos Christeas 2011-10-10 18:54:30 CEST

I think it works,
running python-2.7.1-6.mga1, no hacks AFAICS

Comment 12 Marja Van Waes 2011-10-10 19:23:34 CEST

I understand this bug got fixed. Feel free to reopen when I'm mistaken

Status: NEW => RESOLVED
Resolution: (none) => FIXED