Re: [Libguestfs] [PATCH libnbd] examples: Fix theoretical cookie race in example.

On 7/30/19 5:54 AM, Richard W.M. Jones wrote: We already have option (2) to some extent; libnbd.pod states: The completion callback will be invoked with C<cookie> set to the same value returned by the original API such as C<nbd_aio_pread_callback> (in rare cases, it is possible that the completion callback may fire before the original API has returned). But I also favor option (1); my recent changes to nbdkit to use auto-retire also managed to get rid of the need to rely on the cookie argument. The cookie is still important for clients not using auto-retire, but by getting rid of the cookie parameter to the completion callback, we are forcing the client to do their own mapping of their void* parameter to determine which command is being completed - but that turns out to be the most efficient anyways (both nbdkit and your glib example got that that point without too much difficulty). It's another API/ABI break, but I don't see it as being detrimental (none of our existing completion callbacks are too difficult to keep working if they lose the cookie parameter). -- Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3226 Virtualization: qemu.org | libvirt.org