Fix things so that array_agg_finalfn does not modify or free its input

ArrayBuildState, per trouble report from Merlin Moncure.  By adopting
this fix, we are essentially deciding that aggregate final-functions
should not modify their inputs ever.  Adjust documentation and comments
to match that conclusion.

doc/src/sgml/xaggr.sgml

@@ -1,4 +1,4 @@
-<!-- $PostgreSQL: pgsql/doc/src/sgml/xaggr.sgml,v 1.37 2008/12/28 18:53:54 tgl Exp $ -->
+<!-- $PostgreSQL: pgsql/doc/src/sgml/xaggr.sgml,v 1.38 2009/06/20 18:45:28 tgl Exp $ -->
 
  <sect1 id="xaggr">
   <title>User-Defined Aggregates</title>
@@ -175,10 +175,14 @@ SELECT attrelid::regclass, array_accum(atttypid::regtype)
             (IsA(fcinfo-&gt;context, AggState) ||
              IsA(fcinfo-&gt;context, WindowAggState)))
 </programlisting>
-   One reason for checking this is that when it is true, the first input
+   One reason for checking this is that when it is true for a transition
+   function, the first input
    must be a temporary transition value and can therefore safely be modified
    in-place rather than allocating a new copy.  (This is the <emphasis>only</>
-   case where it is safe for a function to modify a pass-by-reference input.)
+   case where it is safe for a function to modify a pass-by-reference input.
+   In particular, aggregate final functions should not modify their inputs in
+   any case, because in some cases they will be re-executed on the same
+   final transition value.)
    See <literal>int8inc()</> for an example.
   </para>
 

src/backend/executor/nodeWindowAgg.c

@@ -27,7 +27,7 @@
  * Portions Copyright (c) 1994, Regents of the University of California
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/executor/nodeWindowAgg.c,v 1.5 2009/06/11 14:48:57 momjian Exp $
+ *	  $PostgreSQL: pgsql/src/backend/executor/nodeWindowAgg.c,v 1.6 2009/06/20 18:45:28 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -410,9 +410,8 @@ eval_windowaggregates(WindowAggState *winstate)
 	 * need the current aggregate value.  This is considerably more efficient
 	 * than the naive approach of re-running the entire aggregate calculation
 	 * for each current row.  It does assume that the final function doesn't
-	 * damage the running transition value.  (Some C-coded aggregates do that
-	 * for efficiency's sake --- but they are supposed to do so only when
-	 * their fcinfo->context is an AggState, not a WindowAggState.)
+	 * damage the running transition value, but we have the same assumption
+	 * in nodeAgg.c too (when it rescans an existing hash table).
 	 *
 	 * In many common cases, multiple rows share the same frame and hence the
 	 * same aggregate value. (In particular, if there's no ORDER BY in a RANGE

src/backend/utils/adt/array_userfuncs.c

@@ -6,7 +6,7 @@
  * Copyright (c) 2003-2009, PostgreSQL Global Development Group
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/utils/adt/array_userfuncs.c,v 1.30 2009/06/11 14:49:03 momjian Exp $
+ *	  $PostgreSQL: pgsql/src/backend/utils/adt/array_userfuncs.c,v 1.31 2009/06/20 18:45:28 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -537,10 +537,13 @@ array_agg_finalfn(PG_FUNCTION_ARGS)
 	dims[0] = state->nelems;
 	lbs[0] = 1;
 
-	/* Release working state if regular aggregate, but not if window agg */
+	/*
+	 * Make the result.  We cannot release the ArrayBuildState because
+	 * sometimes aggregate final functions are re-executed.
+	 */
 	result = makeMdArrayResult(state, 1, dims, lbs,
 							   CurrentMemoryContext,
-							   IsA(fcinfo->context, AggState));
+							   false);
 
 	PG_RETURN_DATUM(result);
 }

src/backend/utils/adt/arrayfuncs.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/utils/adt/arrayfuncs.c,v 1.157 2009/06/11 14:49:03 momjian Exp $
+ *	  $PostgreSQL: pgsql/src/backend/utils/adt/arrayfuncs.c,v 1.158 2009/06/20 18:45:28 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -4179,9 +4179,21 @@ accumArrayResult(ArrayBuildState *astate,
 		}
 	}
 
-	/* Use datumCopy to ensure pass-by-ref stuff is copied into mcontext */
+	/*
+	 * Ensure pass-by-ref stuff is copied into mcontext; and detoast it too
+	 * if it's varlena.  (You might think that detoasting is not needed here
+	 * because construct_md_array can detoast the array elements later.
+	 * However, we must not let construct_md_array modify the ArrayBuildState
+	 * because that would mean array_agg_finalfn damages its input, which
+	 * is verboten.  Also, this way frequently saves one copying step.)
+	 */
 	if (!disnull && !astate->typbyval)
-		dvalue = datumCopy(dvalue, astate->typbyval, astate->typlen);
+	{
+		if (astate->typlen == -1)
+			dvalue = PointerGetDatum(PG_DETOAST_DATUM_COPY(dvalue));
+		else
+			dvalue = datumCopy(dvalue, astate->typbyval, astate->typlen);
+	}
 
 	astate->dvalues[astate->nelems] = dvalue;
 	astate->dnulls[astate->nelems] = disnull;