"David G. Johnston" <david.g.johnston@gmail.com> writes:
> On Fri, Jul 15, 2022 at 2:06 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Here's a draft patch to fix this. We basically just need to force
>> finish_xact_command() in the same way as we do for transaction control
>> statements. I considered using the same technology as the code uses
>> for transaction control --- that is, statically check for the types of
>> statements that are trouble --- but after reviewing the set of callers
>> of PreventInTransactionBlock() I gave that up as unmaintainable.
> This seems like too narrow a fix though.
I read this, and I have absolutely no idea what you're talking about
or what you concretely want to do differently. If it's just a
documentation question, I agree that I didn't address docs yet.
Probably we do need to put something in the protocol chapter
pointing out that some commands will commit immediately.
I'm not sure I buy your argument that there's a fundamental
difference between simple and extended query protocol in this
area. In simple protocol you can wrap an "implicit transaction"
around several commands by sending them in one query message.
What we've got here is that you can do the same thing in
extended protocol by omitting Syncs. Extended protocol's
skip-till-Sync-after-error behavior is likewise very much like
the fact that simple protocol abandons the rest of the query
string after an error.
regards, tom lane
