Here's a patch for Versions 1 and 2 that fixes the following bug:

When you try to do any UPDATE of the catalog class pg_class, such as
to change ownership of a class, the backend crashes.

This is really two serial bugs: 1) there is a hardcoded copy of the
schema of pg_class in the postgres program, and it doesn't match the
actual class that initdb creates in the database; 2) Parts of postgres
determine whether to pass an attribute value by value or by reference
based on the attbyval attribute of the attribute in class
pg_attribute.  Other parts of postgres have it hardcoded.  For the
relacl[] attribute in class pg_class, attbyval does not match the
hardcoded expectation.

The fix is to correct the hardcoded schema for pg_attribute and to
change the fetchatt macro so it ignores attbyval for all variable
length attributes.  The fix also adds a bunch of logic documentation and
extends genbki.sh so it allows source files to contain such documentation.

--
Bryan Henderson                                    Phone 408-227-6803
San Jose, California

1 files changed, 14 insertions, 2 deletions

diff --git a/src/backend/access/tupmacs.h b/src/backend/access/tupmacs.h
index 9a9bcce3b41..168a4669c3b 100644
--- a/src/backend/access/tupmacs.h
+++ b/src/backend/access/tupmacs.h
@@ -6,7 +6,7 @@
  *
  * Copyright (c) 1994, Regents of the University of California
  *
- * $Id: tupmacs.h,v 1.1.1.1 1996/07/09 06:21:09 scrappy Exp $
+ * $Id: tupmacs.h,v 1.2 1996/08/21 04:25:37 scrappy Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -22,6 +22,12 @@
  * given a AttributeTupleForm and a pointer into a tuple's data
  * area, return the correct value or pointer.
  *
+ * We return a 4 byte (char *) value in all cases.  If the attribute has
+ * "byval" false or has variable length, we return the same pointer
+ * into the tuple data area that we're passed.  Otherwise, we return
+ * the 1, 2, or 4 bytes pointed to by it, properly extended to 4
+ * bytes, depending on the length of the attribute.
+ *
  * note that T must already be properly LONGALIGN/SHORTALIGN'd for
  * this to work correctly.
  *
@@ -30,9 +36,15 @@
  * sign-extension may get weird if you use an integer type that
  * isn't the same size as (char *) for the first cast.  (on the other
  * hand, it's safe to use another type for the (foo *)(T).)
+ *
+ * attbyval seems to be fairly redundant.  We have to return a pointer if
+ * the value is longer than 4 bytes or has variable length; returning the
+ * value would be useless.  In fact, for at least the variable length case,
+ * the caller assumes we return a pointer regardless of attbyval.
+ * I would eliminate attbyval altogether, but I don't know how.  -BRYANH.
  */
 #define fetchatt(A, T) \
- ((*(A))->attbyval \
+ ((*(A))->attbyval && (*(A))->attlen != -1 \
   ? ((*(A))->attlen > sizeof(int16) \
      ? (char *) (long) *((int32 *)(T)) \
      : ((*(A))->attlen < sizeof(int16) \