Hi,
I was looking at the code for bitmap index:
http://archives.postgresql.org/pgsql-hackers/2008-10/msg01691.php
and I couldn't understand why during "bmbuild" (the ambuild call for
bitmap indexes) the code checks for not-ordered ItemPointerData(s).
In which cases the ItemPointerData(s) given by IndexBuildHeapScan
are not in order (when allow_sync=false)?
Leonardo