On Wed, Jul 22, 2009 at 12:22:08PM +0100, Daniel P. Berrange wrote:
On Wed, Jul 22, 2009 at 01:20:57AM -0400, Laine Stump wrote:
If the mac address of an interface is changed, we must either create a new virInterface object in the cache that has the new name/mac, or modify the existing object in the cache. Because the cache is in a hash that's indexed by name only, we can't create a new entry having the same name without destroying the old one first (which isn't a possibility as someone else is already referencing the old one), so we're stuck modifying the existing entry. We have to do that without changing the pointer to the mac though, so we really can only do it if the length of the new mac is equal to or less than the old mac. Otherwise, we have to just bail.
Fixing this problem to properly handle this corner case would require a hash table that used both name and mac as keys, but that would be inefficient (using the existing hash table implementation, which only allows a single key), and probably unnecessary, as the length of mac for any given interface will never change in practice.
Yep, the odds of this error case happening are sooooo small, that giving a fatal error here is more than good enough.
Agreed, patch applied, I only had to add _() to get the message localized, Daniel -- Daniel Veillard | libxml Gnome XML XSLT toolkit http://xmlsoft.org/ daniel@veillard.com | Rpmfind RPM search engine http://rpmfind.net/ http://veillard.com/ | virtualization library http://libvirt.org/