Daniel Veillard wrote:
On Mon, Jul 23, 2007 at 04:00:59PM +0100, Richard W.M. Jones wrote:
Previously[1] I patched libvirt to add the virDomainGetConnect and
virNetworkGetConnect functions. The rationale for these is explained
here[2].
The attached patch makes the Python bindings use these functions instead
of storing the connection object explicitly.
Hum ... The problem was the following:
- you create a connection
- you get a domain pointer from the connection
At the C level, conn->uses is incremented when we get a new domain object.
- you delete the reference from the connection (e.g. by exiting
the method) but keep the domain object
- the connection count of the python object goes down to 0
- _del() is called on the object
- the connection is closed at the C level
conn->uses should still be >= 1 (reflecting the outstanding domain
object), so the connection shouldn't be closed.
- you have a dandling domain python object whose C pointer is dead
Are you sure your patch is safe ?
Having said that, I will now go and construct a test case to see if this
really is safe.
garbage.collection++
Rich.
--
Libvir-list mailing list
Libvir-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/libvir-list