[ZODB-Dev] Race condition in basestorage commit locks?

Christian Theune ct at gocept.com
Fri Oct 7 15:36:02 EDT 2005


Am Freitag, den 07.10.2005, 14:44 -0400 schrieb Tim Peters:
> [Christian Theune]
> > It looks like we have been bitten by a bug in a C-Module for MySQL
> > access. Sometimes it led to a segfault, a halt, or the hang. We can't
> > reproduce it after switching that C-Module anymore.
> 
> Any theory for how that could make all four threads appear to be blocked on
> an attempt to get the commit lock?  Bad code at the C level can cause any
> kind of "impossible" behavior, but there's usually a more-or-less clear
> connection to the impossible symptom observed.  Without such a connection,
> it remains plausible that, e.g., some _other_ bad C-level code is really at
> fault, and switching an irrelevant bit of C code just happened to mask it
> for the time being.

Unfortunately not. Did I mention we had those problems on a production
system? *yuck* No time to play around except of fixing it ... :/

> > I really hate C-Modules for Python right now.
> 
> They're no worse than C code written for other purposes.  I just noticed
> that doing a Google search on
> 
>     mysql segfault
> 
> gives over 200000 hits.  That's about average for C code <wink>.

Right. :) (They especially like to segfault when mysql server and client
libraries don't exactly match in version numbers)

One paper, that is unfortunately rather old, I found a couple of hours
ago and which sounds very intesting to me:

http://systems.cs.uchicago.edu/wad/python.html

Didn't we have those features in Ada already? ;)

Cheers,
Christian

-- 
gocept gmbh & co. kg - schalaunische str. 6 - 06366 koethen - germany
www.gocept.com - ct at gocept.com - phone +49 3496 30 99 112 -
fax +49 3496 30 99 118 - zope and plone consulting and development
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : http://mail.zope.org/pipermail/zodb-dev/attachments/20051007/ee52d891/attachment.bin


More information about the ZODB-Dev mailing list