Re: [HACKERS] The number of bytes is stored in index_size of pgstatindex() ?

I wrote:
> Peter Geoghegan <pg@heroku.com> writes:
>> I think we should change it. It seems like a bug to me.

> Me too.  Is it enough bug-like to be something to back-patch, or should
> we just change it in HEAD?

Actually, there's a significantly worse bug here: I just realized that the
page type tests are done in the wrong order.  A deleted page that was
formerly a leaf will be reported as though it was a live leaf page,
because both the BTP_LEAF and BTP_DELETED flags are set for such a page.

It looks like this was done correctly to begin with, and I broke it in
d287818eb514d431b1a68e1f3940cd958f82aa34.  Not sure what I was thinking :-(

Anyway, I think that puts the final nail in the coffin of the idea that
the current code's behavior is sane enough to preserve.  I think we should
fix all these things and back-patch 'em all.

            regards, tom lane