Horton

Re: libpq copy error handling busted

Thomas Munro <thomas.munro@gmail.com>

View thread View on postgr.es
From: Thomas Munro <thomas.munro@gmail.com>
To: Oleksandr Shulgin <oleksandr.shulgin@zalando.de>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, Andres Freund <andres@anarazel.de>, pgsql-hackers <pgsql-hackers@postgresql.org>
Date: 2020-06-04T09:04:59Z
Lists: pgsql-hackers 
On Thu, Jun 4, 2020 at 6:22 PM Oleksandr Shulgin
<oleksandr.shulgin@zalando.de> wrote:
> On Thu, Jun 4, 2020 at 5:37 AM Thomas Munro <thomas.munro@gmail.com> wrote:
>> Here's what I tested.  First, I put this into pgdata/postgresql.conf:

> Would it be feasible to capture this in a sort of a regression (TAP?) test?

If I'm remembering correctly, it wouldn't work on Windows.  After
you've had an error sending to a socket, you can't receive (even if
there was something sent to you earlier).  At least that's how it
seemed from the experiments on that other thread.  The other problem
is that it requires inserting a sleep into the code...

Commits

  1. Try to read data from the socket in pqSendSome's write_failed paths.

  2. Rethink definition of cancel.c's CancelRequested flag.

  3. Restructure libpq's handling of send failures.