Re: setting per-database/role parameters checks them against wrong context

On Fri, Sep 28, 2012 at 7:10 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> Peter Eisentraut <peter_e(at)gmx(dot)net> writes:
>> Example:
>> create temporary table foo (a int);
>> insert into foo values (1);
>> alter role peter set temp_buffers = 120;
>> ERROR: 22023: invalid value for parameter "temp_buffers": 120
>> DETAIL: "temp_buffers" cannot be changed after any temporary tables
>> have been accessed in the session.
>
>> Another example:
>
>> set log_statement_stats = on;
>> alter role peter set log_parser_stats = on;
>> ERROR: 22023: invalid value for parameter "log_parser_stats": 1
>> DETAIL: Cannot enable parameter when "log_statement_stats" is true.
>
>> Another example is that in <=9.1, ALTER DATABASE ... SET search_path
>> would check the existence of the schema in the current database, but
>> that was done away with in 9.2.
>
>> The first example could probably be fixed if check_temp_buffers() paid
>> attention to the GucSource, but in the second case and in general there
>> doesn't seem to be a way to distinguish "assigning per-database setting"
>> and "enacting per-database setting" as a source.
>
>> Ideas?
>
> Perhaps instead of trying to solve the problem as stated, it would be
> more useful to put the effort into getting rid of context-sensitive
> restrictions on GUC settings. Neither of the examples above seem
> particularly essential - they are just protecting incomplete
> implementations.

The check_temp_buffers() problem seems like a regression and blocks us
from upgrading to 9.2. The use case are functions that set
temp_buffers and occasionally are called in a series from a parent
session. The work around is... a lot of work.

I'd appreciate it if we could fix the temp_buffers issue, and I
support getting rid of context-sensitive restrictions. :)

-selena

--
http://chesnok.com