On 10/18/2016 04:13 PM, Tom Lane wrote:
> Magnus Hagander <magnus@hagander.net> writes:
>> On Tue, Oct 18, 2016 at 1:00 AM, Vladimir Sitnikov <
>> sitnikov.vladimir@gmail.com> wrote:
>>> The test executes "select pg_sleep(10)" and tries to cancel it. In recent
>>> master builds, cancel seems to be ignored, and the statement lasts for 10
>>> seconds.
>
>> My guess is it's related to this:
>> https://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;h=9e083fd4683294f41544e6d0d72f6e258ff3a77c
>> That's certainly not intended to break things, but that was changed on Oct
>> 17 and it relates to cancel keys.
>> What platform does the postgres server run on? Can can you check if query
>> cancel works on libpq or if it's completely broken?
>
> I can confirm that query cancel is broken in HEAD on RHEL6.
>
> regression=# select pg_sleep(10);
> ^CCancel request sent
> ... nothing happens for the balance of the 10 seconds ...
> regression=#
>
> There's a smoking gun in the postmaster log:
>
> 2016-10-18 09:10:34.547 EDT [18502] LOG: wrong key in cancel request for process 18491
Ok, I've reverted that commit for now. It clearly needs more thought,
because of this, and the pademelon failure discussed on the other thread.
- Heikki
