On Tue, Dec 04, 2007 at 03:50:46PM +0000, Richard W.M. Jones wrote:
Daniel Veillard wrote:
On Tue, Dec 04, 2007 at 12:57:39PM +0000, Richard W.M. Jones wrote:
Daniel P. Berrange wrote:
virConnectPtr virConnectOpenAuth (const char *name, virConnectAuthPtr auth, int flags); I'm a fan of callers passing in the size of the structure (as they see it). Allows the structure to be expanded in future, and if done right can allow both forwards and backwards compatibility.
cf: http://www.libvirt.org/html/libvirt-libvirt.html#virDomainInterfaceStats
Hum, honnestly, that's not my preferred way. If you really think there should be room for expansion, I would either: - add a version number to the structure and allocator/destructor functions as part of the API (prefered)
Hmmm ... what if they forget to set it? Now we need an init function for the structure.
yes that's why I suggested an allocator, it would set the version.
- add padding at tyhe end of the structure which could allow a future growth
This incurs an unnecessary memory penalty always, and limits us to a particular maximum size which we must choose now when we don't know what extra fields we might add.
I don't like this much either, but I think the GNOME guys did this for libxml2 I usually provide allocator/deallocators.
adding the size of the structure as the argument moves the complexity away from the library implementor to the library user,
Well, it means they must append "sizeof (*auth)" to their list of arguments, which is hardly a lot of complexity. There _is_ much more
until they just add a numeric value ... and that will break just as well. trying to minimize the cases where the API user can do the wrong thing is an important goal :-) . There that's something which seems to work but will break later, that's why I prefer an allocator if we plan to change sizes.
Remember also the need for forwards compatibility: program which was compiled against an old version of the library, dynamically links to a new version of the library. If the structure has increased in size in the meantime, then having the caller pass the size means that the library can do the right thing and only fill in the first fields.
The version + allocator have that property too, except it's the library code which takes care of things. Daniel -- Red Hat Virtualization group http://redhat.com/virtualization/ Daniel Veillard | virtualization library http://libvirt.org/ veillard@redhat.com | libxml GNOME XML XSLT toolkit http://xmlsoft.org/ http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/