On 01/27/15 at 09:22am, Simon Horman wrote: > Hi Ben, Hi Thomas, > > On Mon, Jan 26, 2015 at 11:19:53AM -0800, Ben Pfaff wrote: > > On Mon, Jan 26, 2015 at 02:44:07PM +0100, Thomas Graf wrote: > > > list_moved() not handling the list_empty() case is somewhat rude. > > > Why not just handle that special case and call list_init() in > > > list_moved() instead? I realize it's currently documented to fail in > > > list_moved() but I don't see any side effect in handling this properly > > > because any caller doing so right now would be buggy. > > > > I'd prefer to handle that case in list_moved() but I don't know a safe > > way to detect it, that is, a way that doesn't try to read from > > possibly freed memory. > > Likewise. > > I had an idea to detect it using something like this, but > it breaks list_is_short() and moreover changes the scheme > of empty lists.
The only really clean solution I could come up with is to replace the realloc() with an explicit alloc() + memcpy() + free() and do a list_move() before the free. That would allow to get rid of list_moved() completely. Not really better than what Simon proposes though so I have no objection to Simon's patch. Acked-by: Thomas Graf <tg...@noironetworks.com> _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
