Fix a passel of problems with incorrect calls to typinput and typoutput

functions, which would lead to trouble with datatypes that paid attention
to the typelem or typmod parameters to these functions.  In particular,
incorrect code in pg_aggregate.c explains the platform-specific failures
that have been reported in NUMERIC avg().

1 files changed, 4 insertions, 5 deletions

diff --git a/src/backend/bootstrap/bootstrap.c b/src/backend/bootstrap/bootstrap.c
index 8e9f125996d..306b983370b 100644
--- a/src/backend/bootstrap/bootstrap.c
+++ b/src/backend/bootstrap/bootstrap.c
@@ -7,7 +7,7 @@
  * Copyright (c) 1994, Regents of the University of California
  *
  * IDENTIFICATION
- *	  $Header: /cvsroot/pgsql/src/backend/bootstrap/bootstrap.c,v 1.77 2000/01/15 02:59:27 petere Exp $
+ *	  $Header: /cvsroot/pgsql/src/backend/bootstrap/bootstrap.c,v 1.78 2000/01/15 22:43:20 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -680,11 +680,10 @@ InsertOneValue(Oid objectid, char *value, int i)
 		values[i] = fmgr(ap->am_typ.typinput,
 						 value,
 						 ap->am_typ.typelem,
-						 -1);	/* shouldn't have char() or varchar()
-								 * types during boostrapping but just to
-								 * be safe */
+						 -1);
 		prt = fmgr(ap->am_typ.typoutput, values[i],
-				   ap->am_typ.typelem);
+				   ap->am_typ.typelem,
+				   -1);
 		if (!Quiet)
 			printf("%s ", prt);
 		pfree(prt);