On Wed, Sep 30, 2015 at 09:47:16AM -0700, Joe Stringer wrote: > I think the main change I was looking for in this series is to support > >64bit values by switching the value parsing to use parse_int_string() > instead of strtoull() directly, using the mf field's expected length > to figure out how long the string is allowed to be.
OK, doing that makes sense. > The other matter was that there seems to be a discrepancy between the > error checking code in set_field vs reg_load, for instance checking > whether the field is read-only, or passing it through > mf_is_value_valid(). Again, that doesn't require the current approach > to address. Some of that comes down to backward compatibility, since the requirements imposed by OpenFlow on set_field have subtle differences from those imposed by OVS on reg_load. That doesn't have to be imposed on the basis of how the action is written in text form, I guess, but any given load or set_field has to satisfy either one or the other set of requirements. _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
