Fix pg_stat_statements for EXEC_BACKEND case.

We should not try to load old statistics when re-attaching to existing
shared memory.  Per bug #4941.

Itagaki Takahiro
This commit is contained in:
Tom Lane 2009-07-27 03:34:40 +00:00
parent 8835d63b27
commit e4b9e65393
1 changed files with 2 additions and 2 deletions

View File

@ -14,7 +14,7 @@
* Copyright (c) 2008-2009, PostgreSQL Global Development Group * Copyright (c) 2008-2009, PostgreSQL Global Development Group
* *
* IDENTIFICATION * IDENTIFICATION
* $PostgreSQL: pgsql/contrib/pg_stat_statements/pg_stat_statements.c,v 1.3 2009/06/11 14:48:51 momjian Exp $ * $PostgreSQL: pgsql/contrib/pg_stat_statements/pg_stat_statements.c,v 1.4 2009/07/27 03:34:40 tgl Exp $
* *
*------------------------------------------------------------------------- *-------------------------------------------------------------------------
*/ */
@ -332,7 +332,7 @@ pgss_shmem_startup(void)
* Note: we don't bother with locks here, because there should be no other * Note: we don't bother with locks here, because there should be no other
* processes running when this is called. * processes running when this is called.
*/ */
if (!pgss_save) if (found || !pgss_save)
return; return;
file = AllocateFile(PGSS_DUMP_FILE, PG_BINARY_R); file = AllocateFile(PGSS_DUMP_FILE, PG_BINARY_R);