Re: postmaster core ( finally I have it )

From: "Tom Lane" <tgl@sss.pgh.pa.us>
> "Mendola Gaetano" <mendola@bigfoot.com> writes:
> > The process killed made always the same select ( with different
> > id_package ):
> 
> > SELECT id_publisher, publisher_name, id_package, package_name
> > FROM v_psl_package_info
> > WHERE id_package = 177;
> 
> > (gdb) where
> > #0  0x08171fdd in RelationBuildRuleLock ()
> > #1  0x081722ab in RelationBuildDesc ()
> > #2  0x0817311c in RelationIdGetRelation ()
> > #3  0x08077b95 in relation_open ()
> > #4  0x08077d24 in heap_openrv ()
> > #5  0x080b1015 in addRangeTableEntry ()
> > #6  0x080ab6de in transformTableEntry ()
> > #7  0x080ab942 in transformFromClauseItem ()
> > #8  0x080ab321 in transformFromClause ()
> > #9  0x0809f639 in transformSelectStmt ()
> > #10 0x0809dcc4 in transformStmt ()
> > #11 0x0809da06 in parse_analyze ()
> 
> Hmm.  RelationBuildRuleLock is concerned with reading the rule
> information for a table; specifically, it's evidently trying to
> read the rules for v_psl_package_info.  I suspect some form of
> data corruption in the pg_rewrite row(s) for this table.  Do you
> see any misbehavior when you do
> 
> select * from pg_rewrite where ev_class = 'v_psl_package_info'::regclass

All seems good.

> In particular, are any of the columns NULL in these rows?

I did:
select * 
from pg_rewrite 
where ev_class = 'v_psl_package_info'::regclass and     ( rulename is null or       ev_class is null or       ev_attr
isnull or       ev_type is null or       is_instead is null or       ev_qual is null or       ev_action is null );
 


and I had:  No rows.



v_psl_package_info is a view, shall we look in another direction ?




thank you 
Gaetano