Quote Originally Posted by billiauk
Interesting this. Same is true on Windows. At first I thought it could be merging, but I see that Padders included an ORDER BY, presumably for this very reason ( i.e. force no merge ).

The only difference in 10g with this that I can see is the FAST DUAL optimisation. I can't see anything else in trace. It could be a bug of course. It doesn't appear to be anything to do with the PLSQL_OPTIMIZE_LEVEL either - setting this to 0 made no difference.
Sorry, didn't see this before posting the thing about the context suffering the same fate.

Cheers

Tim...