Re: FOR KEY LOCK foreign keys

Noah Misch wrote:

> With this patch in its final form, I have completed 180+ suite runs
> without a failure.

The attached patch allows the tests to pass when
default_transaction_isolation is stricter than 'read committed'.
This is a slight change from the previously posted version of the
files (because of a change in the order of statements, based on the
timeouts), and in patch form this time.

Since `make installcheck-world` works at all isolation level
defaults, as do all previously included isolation tests, it seems
like a good idea to keep this up. It will simplify my testing of SSI
changes, anyway.

-Kevin

On Sat, Jul 16, 2011 at 01:03:31PM -0500, Kevin Grittner wrote:
> Noah Misch wrote:
>
> > With this patch in its final form, I have completed 180+ suite runs
> > without a failure.
>
> The attached patch allows the tests to pass when
> default_transaction_isolation is stricter than 'read committed'.
> This is a slight change from the previously posted version of the
> files (because of a change in the order of statements, based on the
> timeouts), and in patch form this time.
>
> Since `make installcheck-world` works at all isolation level
> defaults, as do all previously included isolation tests, it seems
> like a good idea to keep this up. It will simplify my testing of SSI
> changes, anyway.

This does seem sensible. Thanks.

--
Noah Misch http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services

"Kevin Grittner" <Kevin(dot)Grittner(at)wicourts(dot)gov> wrote:
> Noah Misch wrote:
>
>> With this patch in its final form, I have completed 180+ suite
>> runs without a failure.
>
> The attached patch allows the tests to pass when
> default_transaction_isolation is stricter than 'read committed'.

Without these two patches the tests fail about one time out of three
on my machine at the office at the 'read committed' transaction
isolation level, and all the time at stricter levels. On my machine
at home I haven't seen the failures at 'read committed'. I don't
know if this is Intel (at work) versus AMD (at home) or what.

With both Noah's patch and mine I haven't yet seen a failure in
either environment, with a few dozen tries..

-Kevin

Excerpts from Kevin Grittner's message of sáb jul 16 14:03:31 -0400 2011:
> Noah Misch wrote:
>
> > With this patch in its final form, I have completed 180+ suite runs
> > without a failure.
>
> The attached patch allows the tests to pass when
> default_transaction_isolation is stricter than 'read committed'.
> This is a slight change from the previously posted version of the
> files (because of a change in the order of statements, based on the
> timeouts), and in patch form this time.

Thanks, applied. Sorry for the delay.

--
Álvaro Herrera <alvherre(at)commandprompt(dot)com>
The PostgreSQL Company - Command Prompt, Inc.
PostgreSQL Replication, Consulting, Custom Development, 24x7 support

Alvaro Herrera <alvherre(at)commandprompt(dot)com> wrote:
> Excerpts from Kevin Grittner's message:
>> Noah Misch wrote:
>>
>>> With this patch in its final form, I have completed 180+ suite
>>> runs without a failure.
>>
>> The attached patch allows the tests to pass when
>> default_transaction_isolation is stricter than 'read committed'.
>> This is a slight change from the previously posted version of the
>> files (because of a change in the order of statements, based on
>> the timeouts), and in patch form this time.
>
> Thanks, applied. Sorry for the delay.

My patch was intended to supplement Noah's patch here:

http://archives.postgresql.org/pgsql-hackers/2011-07/msg00867.php

Without his patch, there is still random failure on my work machine
at all transaction isolation levels.

-Kevin

Excerpts from Kevin Grittner's message of mar jul 19 13:49:53 -0400 2011:
> Alvaro Herrera <alvherre(at)commandprompt(dot)com> wrote:
> > Excerpts from Kevin Grittner's message:
> >> Noah Misch wrote:
> >>
> >>> With this patch in its final form, I have completed 180+ suite
> >>> runs without a failure.
> >>
> >> The attached patch allows the tests to pass when
> >> default_transaction_isolation is stricter than 'read committed'.
> >> This is a slight change from the previously posted version of the
> >> files (because of a change in the order of statements, based on
> >> the timeouts), and in patch form this time.
> >
> > Thanks, applied. Sorry for the delay.
>
> My patch was intended to supplement Noah's patch here:

I'm aware of that, thanks. I'm getting that one in too, shortly.

--
Álvaro Herrera <alvherre(at)commandprompt(dot)com>
The PostgreSQL Company - Command Prompt, Inc.
PostgreSQL Replication, Consulting, Custom Development, 24x7 support