Re: Signed vs. Unsigned (some)

On Fri, Jul 02, 2021 at 12:09:23PM +0200, Peter Eisentraut wrote:
> On 16.06.21 10:48, Peter Eisentraut wrote:
> > On 15.06.21 10:17, Kyotaro Horiguchi wrote:
> > > The definitions are not ((type) -1) but ((type) 0xFFFFFFFF) so
> > > actually they might be different if we forget to widen the constant
> > > when widening the types.  Regarding to the compiler behavior, I think
> > > we are assuming C99[1] and C99 defines that -1 is converted to
> > > Uxxx_MAX. (6.3.1.3 Singed and unsigned integers)
> > > 
> > > I'm +0.2 on it.  It might be worthwhile as a matter of style.
> > 
> > I think since we have the constants we should use them.
> 
> I have pushed the InvalidBucket changes.
> 
> The use of InvalidBlockNumber with vac_update_relstats() looks a bit fishy
> to me.  We are using in the same call 0 as the default for
> num_all_visible_pages, and we generally elsewhere also use 0 as the starting
> value for relpages, so it's not clear to me why it should be -1 or
> InvalidBlockNumber here.  I'd rather leave it "slightly wrong" for now so it
> can be checked again.

There's two relevant changes:

|commit 3d351d916b20534f973eda760cde17d96545d4c4
|Author: Tom Lane <tgl@sss.pgh.pa.us>
|Date:   Sun Aug 30 12:21:51 2020 -0400
|
|    Redefine pg_class.reltuples to be -1 before the first VACUUM or ANALYZE.

|commit 0e69f705cc1a3df273b38c9883fb5765991e04fe
|Author: Alvaro Herrera <alvherre@alvh.no-ip.org>
|Date:   Fri Apr 9 11:29:08 2021 -0400
|
|    Set pg_class.reltuples for partitioned tables

3d35 also affects partitioned tables, and 0e69 appears to do the right thing by
preserving relpages=-1 during auto-analyze.

Note that Alvaro's commit message and comment refer to relpages, but should
have said reltuples - comment fixed at 7ef8b52cf.

-- 
Justin