On 2017-03-30 19:40:08 +0100, Simon Riggs wrote:
> On 30 March 2017 at 18:16, Andres Freund <andres@anarazel.de> wrote:
>
> >> /*
> >> * Each page of XLOG file has a header like this:
> >> */
> >> -#define XLOG_PAGE_MAGIC 0xD097 /* can be used as WAL version indicator */
> >> +#define XLOG_PAGE_MAGIC 0xD100 /* can be used as WAL version indicator */
> >
> > We normally only advance this by one, it's not tied to the poistgres version.
>
> That was my addition. I rounded it up cos this is release 10. No biggie.
We'll probably upgrade that more than once again this release...
> (Poistgres? Is that the Manhattan spelling?)
Tiredness spelling ;)
> We've been redesigning the mechanisms for 2 years now, so it seems
> unlikely that further redesign can be required.
I don't think that's true *at all* - the mechanism previously
fundamentally different.
The whole topic has largely seen activity shortly before the code
freeze, both last time round and now. I don't think it's surprising
that it thus doesn't end up being ready.
> If it is required,
> this patch is fairly low touch and change is possible later,
> incremental development etc. As regards overhead, this adds a small
> amount of time to a background process executed every 10 secs,
> generates no new WAL records.
>
> So I don't see any reason not to commit this feature, after the minor
> corrections.
It doesn't have any benefit on its own, the locking model doesn't seem
fully there. I don't see much reason to get this in before the release.
- Andres