Обсуждение: Failed assertion, CVS head

Hi,

I am using a version of PostgreSQL compiled from a CVS update of yesterday,
and compiled with  make clean all  make install

One client connection to the database doing routine and low volume population
scripts (using schemas) After several normal runs of the population script, a
run caused the following trap, and an associated backend halt.

TRAP: FailedAssertion("!(((ntp)->t_data)->t_infomask & 0x0010)", File: 
"catcache.c", Line: 1728)

I know this is not a lot of data for a determination of cause, but I have only
had a single instance of this happening, and thought it should at least be
documented.

Regards,
Grant

It's happened again, and in both cases seems to be on a call to
VACUUM FULL

Grant Finnemore wrote:
> Hi,
> 
> I am using a version of PostgreSQL compiled from a CVS update of yesterday,
> and compiled with
>   make clean all
>   make install
> 
> One client connection to the database doing routine and low volume 
> population
> scripts (using schemas) After several normal runs of the population 
> script, a
> run caused the following trap, and an associated backend halt.
> 
> TRAP: FailedAssertion("!(((ntp)->t_data)->t_infomask & 0x0010)", File: 
> "catcache.c", Line: 1728)
> 
> I know this is not a lot of data for a determination of cause, but I 
> have only
> had a single instance of this happening, and thought it should at least be
> documented.
> 
> Regards,
> Grant
> 
> ---------------------------(end of broadcast)---------------------------
> TIP 6: Have you searched our list archives?
> 
>               http://archives.postgresql.org
> 

Grant Finnemore <grantf@guruhut.co.za> writes:
> TRAP: FailedAssertion("!(((ntp)->t_data)->t_infomask & 0x0010)", File: 
> "catcache.c", Line: 1728)

This seems moderately impossible :-(.  Did you get a core dump?  If so
please provide a stack backtrace.
        regards, tom lane

I'm afraid that I did not get a core dump. Sorry.

My normal configure includes both debug and cassert - is there anything
else I should set to ensure core dumps are generated?

Regards,
Grant

Tom Lane wrote:
> Grant Finnemore <grantf@guruhut.co.za> writes:
> 
>>TRAP: FailedAssertion("!(((ntp)->t_data)->t_infomask & 0x0010)", File: 
>>"catcache.c", Line: 1728)
> 
> 
> This seems moderately impossible :-(.  Did you get a core dump?  If so
> please provide a stack backtrace.
> 
>             regards, tom lane
> 
> ---------------------------(end of broadcast)---------------------------
> TIP 1: subscribe and unsubscribe commands go to majordomo@postgresql.org
> 

Grant Finnemore <grantf@guruhut.co.za> writes:
> I'm afraid that I did not get a core dump. Sorry.
> My normal configure includes both debug and cassert - is there anything
> else I should set to ensure core dumps are generated?

Check "ulimit -c" in the postmaster's environment.

Personally I always put "ulimit -c unlimited" into the postmaster start
script. 
        regards, tom lane

Ok, will do. Thanks.

Tom Lane wrote:
> Grant Finnemore <grantf@guruhut.co.za> writes:
> 
>>I'm afraid that I did not get a core dump. Sorry.
>>My normal configure includes both debug and cassert - is there anything
>>else I should set to ensure core dumps are generated?
> 
> 
> Check "ulimit -c" in the postmaster's environment.
> 
> Personally I always put "ulimit -c unlimited" into the postmaster start
> script. 
> 

Some time ago, Grant Finnemore <grantf@guruhut.co.za> wrote:
> TRAP: FailedAssertion("!(((ntp)->t_data)->t_infomask & 0x0010)", File: 
> "catcache.c", Line: 1728)

I think we finally figured out where this came from:

http://archives.postgresql.org/pgsql-bugs/2004-12/msg00128.php
        regards, tom lane