"Akira Kurosawa" <kurosawa-akira@mxc.nes.nec.co.jp> writes:
> This problem happens because "pgstat_info" was not initialized in the
> function
> "pgstat_initstats" when relkind = 'S'. (pgstat_info = NULL).
Looks like my oversight :-(. Patch applied, thanks for the report!
Interestingly, it seems that this causes nextval() to consistently
increment the sequence's blocks-read or blocks-hit count, which
according to my testing it did not always do pre-8.3. The commit
message for the patch that created this bug mentions
And fix a few
corner cases in which counts might be missed because the relation's
pgstat_info pointer hadn't been set.
so I guess this is a side-effect of that.
regards, tom lane