* jamal [ref] 2005-05-27 09:58

On Fri, 2005-27-05 at 14:50 +0200, Thomas Graf wrote:

quoted

 
-#define NLMSG_PUT(skb, pid, seq, type, len) \
+#define NLMSG_NEW(skb, pid, seq, type, len, flags) \

How about introducing NLMSG_NEW and leave NLMSG_PUT where it is?
This way we could do incremental updates. If there are not too many of
those NLMSG_PUT around, then it shouldnt be an issue.

That is what the patch does but the diff looks a bit irritating.
1) Introduce NLMSG_NEW to take flags
2) Make NLMSG_PUT call NLMSG_NEW with flags==0
3) Change __nlmsg_put to take flags argument

The first change would be definetely on top of the current patch i sent
which is well deserved.
Lets have Dave swallow my patch then either you or i could make those
changes. Sounds reasonable?

Sure.