> I think my original patch about NOT NULL constraint allows this type of
> construction. However, thinks have changed in the parser since I made
> this patch, and now seems that a construction like
> CREATE TABLE a (b int constraint not_null NOT NULL)
> is now valid (I only saw gram.y - maybe I'm wrong). I can make the
> patches to allow NULL, but I only want a litte help: is the name of this
> type of constraint beeing saved anywere?
Yes, I think so.