Re: Confine vacuum skip logic to lazy_scan_skip

On Fri, Feb 28, 2025 at 2:29 PM Thomas Munro <thomas.munro@gmail.com> wrote:
> On Fri, Feb 28, 2025 at 11:58 AM Melanie Plageman
> <melanieplageman@gmail.com> wrote:
> > On Thu, Feb 27, 2025 at 1:08 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> > > I wonder if it'd be a good idea to add something like
> > >
> > >                 Assert(stream->distance == 1);
> > >                 Assert(stream->pending_read_nblocks == 0);
> > >                 Assert(stream->per_buffer_data_size == 0);
> > > +               Assert(per_buffer_data == NULL);
> > >
> > > in read_stream_next_buffer.  I doubt that this will shut Coverity
> > > up, but it would help to catch caller coding errors, i.e. passing
> > > a per_buffer_data pointer when there's no per-buffer data.
> >
> > I think this is a good stopgap. I was discussing adding this assert
> > off-list with Thomas and he wanted to detail his more ambitious plans
> > for type safety improvements in the read stream API. Less on the order
> > of a redesign and more like a separate read_stream_next_buffer()s for
> > when there is per buffer data and when there isn't. And a by-value and
> > by-reference version for the one where there is data.
>
> Here's what I had in mind.  Is it better?

Here's a slightly better one.  I think when you use
read_stream_get_buffer_and_value(stream, &value), or
read_stream_put_value(stream, space, value), then we should assert
that sizeof(value) strictly matches the available space, as shown.  But,
new in v2, if you use read_stream_get_buffer_and_pointer(stream,
&pointer), then sizeof(*pointer) should only have to  be <= the
storage space, not ==, because someone might plausibly want to make
per_buffer_data_size variable at runtime (ie decide when they
construct the stream), and then be able to retrieve a pointer to the
start of a struct with a flexible array or something like that.  In v1
I was just trying to assert that it was a
pointer-to-a-pointer-to-something and no more (in a confusing
compile-time assertion), but v2 is simpler, and is happy with a
pointer to a pointer to something that doesn't exceed the space
(run-time assertion).