Bruce Momjian wrote:
>
> OK, I have looked at this, but can't figure out how to fix the many
> initializations of ScanKeyData. Can someone who understands this please
> submit a patch for all these initializations so we can stop these vacuum
> analyze reports?
>
> Vadim has found the problem, but we need someone to properly fix it.
Just apply my patch to stop "analyze-problem" reports (sorry, I didn't it).
As for other (possible) places, note that ScanKeyEntryInitialize()
initializes sk_func.fn_oid and so we have to worry about hard-coded
initializations only (when ScanKeyEntryInitialize() is not called).
Vadim