Обсуждение: pgsql: Change timeline field of IDENTIFY_SYSTEM to int8
Change timeline field of IDENTIFY_SYSTEM to int8 It was int4, but in the other replication commands, timelines are returned as int8. Reviewed-by: Nathan Bossart <nathandbossart@gmail.com> Discussion: https://www.postgresql.org/message-id/flat/7e4fdbdc-699c-4cd0-115d-fb78a957fc22@enterprisedb.com Branch ------ master Details ------- https://git.postgresql.org/pg/commitdiff/ec40f3422412cfdc140b5d3f67db7fd2dac0f1e2 Modified Files -------------- doc/src/sgml/protocol.sgml | 2 +- src/backend/replication/walsender.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
Peter Eisentraut <peter@eisentraut.org> writes:
> Change timeline field of IDENTIFY_SYSTEM to int8
Surely this patch is far from complete?
To start with, just a few lines down in IdentifySystem() the column
is filled using Int32GetDatum not Int64GetDatum. I will get some
popcorn and await the opinions of the 32-bit buildfarm animals.
But what about whatever code is reading the output? And what if
that code isn't v16? I can't believe that we can make a wire
protocol change as summarily as this.
regards, tom lane
I wrote: > To start with, just a few lines down in IdentifySystem() the column > is filled using Int32GetDatum not Int64GetDatum. I will get some > popcorn and await the opinions of the 32-bit buildfarm animals. Didn't need to wait long: https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=florican&dt=2022-07-04%2005%3A39%3A50 The other questions still stand. regards, tom lane
On Mon, Jul 04, 2022 at 01:55:13AM -0400, Tom Lane wrote: > Peter Eisentraut <peter@eisentraut.org> writes: > > Change timeline field of IDENTIFY_SYSTEM to int8 > > Surely this patch is far from complete? Yeah.. > But what about whatever code is reading the output? And what if > that code isn't v16? I can't believe that we can make a wire > protocol change as summarily as this. Assuming that one reaches a timeline of 2 billion, this change would make the TLI consumption of the client safe to signedness. But why is it safe to do a protocol change when running IDENTIFY_SYSTEM? We've been very strict to maintain compatibility for any protocol change, hence why should the replication protocol be treated differently? -- Michael