Обсуждение: BUG #9074: prepared_xacts test freezes

The following bug has been logged on the website:

Bug reference:      9074
Logged by:          Marco Atzeri
Email address:      marco.atzeri@gmail.com
PostgreSQL version: 9.3.2
Operating system:   cygwin
Description:

going step by step, the freezing point is:

-- pxtest3 should be locked because of the pending DROP
set statement_timeout to 2000;
SELECT * FROM pxtest3;
instead of
--------------------------------
ERROR:  canceling statement due to statement timeout
-----------------------------------
the SELECT never returns.

marco.atzeri@gmail.com writes:
> going step by step, the freezing point is:

> -- pxtest3 should be locked because of the pending DROP
> set statement_timeout to 2000;
> SELECT * FROM pxtest3;
> instead of
> --------------------------------
> ERROR:  canceling statement due to statement timeout
> -----------------------------------
> the SELECT never returns.

Something broken in SIGALRM signal handling, seems like.  But in this
case, the code *does* work on older Cygwin versions, so I think it's
incumbent on you to figure out why newer Cygwin is failing.

            regards, tom lane

On 03/02/2014 07:31, Tom Lane wrote:
> marco.atzeri@gmail.com writes:
>> going step by step, the freezing point is:
>
>> -- pxtest3 should be locked because of the pending DROP
>> set statement_timeout to 2000;
>> SELECT * FROM pxtest3;
>> instead of
>> --------------------------------
>> ERROR:  canceling statement due to statement timeout
>> -----------------------------------
>> the SELECT never returns.
>
> Something broken in SIGALRM signal handling, seems like.  But in this
> case, the code *does* work on older Cygwin versions, so I think it's
> incumbent on you to figure out why newer Cygwin is failing.
>
>             regards, tom lane

but an indication of where to look for and about postgresql expectation
will help to make a test case for SIGALRM stand alone

Thanks
Marco

On 2/3/2014 7:31 AM, Tom Lane wrote:
> marco.atzeri@gmail.com writes:
>> going step by step, the freezing point is:
>
>> -- pxtest3 should be locked because of the pending DROP
>> set statement_timeout to 2000;
>> SELECT * FROM pxtest3;
>> instead of
>> --------------------------------
>> ERROR:  canceling statement due to statement timeout
>> -----------------------------------
>> the SELECT never returns.
>
> Something broken in SIGALRM signal handling, seems like.  But in this
> case, the code *does* work on older Cygwin versions, so I think it's
> incumbent on you to figure out why newer Cygwin is failing.
>
>             regards, tom lane

for the archives, the cygwin bug was identified and solved.

postgresql-9.4.0

      prepared_xacts           ... ok

Regards
Marco