Oleg Mayevskiy <oleg(dot)mayevskiy(at)s2002(dot)tu-chemnitz(dot)de> writes:
> it seems to be a simple problem, but it is not.
Yes it is ... you're expecting the RI triggers to fire during the
plpgsql function, but in fact they fire at completion of the outer
statement that called the plpgsql function.
There's been some debate about whether this is really the most
desirable behavior, but that's how it is at the moment.
regards, tom lane