issue2550549 Postgres error on message templating

Exception gets thrown and not captured if nodeid is too large on
postgres. Added a check in rdbms_common layer that max nodeid is <
2^31 -1. This is also the limit for INTEGER on MySql. SqlLite can
support up to 2^63-1, but I chose the smallest common denominator.

Large nodeid should now returns no such id rather than a stack trace.

Patch idea from: martin.v.loewis. (John Rouillard)

This fixes the reason for patches:

   https://hg.python.org/tracker/roundup/rev/1548763e8273

and

   https://hg.python.org/tracker/roundup/rev/eb3be57f1ef2

Author: rouilj

CHANGES.txt

@@ -113,6 +113,12 @@ Fixed:
   the codecs.
 - issue2550823 improve mailgw logging for node creation errors.
   Patch by r.david.murray (applied by John Rouillard).
+- issue2550549 Postgres error on message templating
+  Exception gets thrown and not captured if nodeid is too large
+  on postgres. Added a check in rdbms_common layer that max nodeid
+  is < 2^31 -1. Large nodeid now return no such id error upstream.
+  Patch idea from: martin.v.loewis. (John Rouillard)
+
 
 2016-01-11: 1.5.1
 

roundup/backends/rdbms_common.py

@@ -1228,6 +1228,15 @@ def destroynode(self, classname, nodeid):
     def hasnode(self, classname, nodeid):
         """ Determine if the database has a given node.
         """
+        # nodeid (aka link) is sql type INTEGER.  max positive value
+        # for INTEGER is 2^31-1 for Postgres and MySQL. The max
+        # positive value for SqLite is 2^63 -1, so arguably this check
+        # needs to adapt for the type of the RDBMS. For right now,
+        # choose lowest common denominator.
+        if int(nodeid) >= 2**31:
+            # value out of range return false
+            return 0
+
         # If this node is in the cache, then we do not need to go to
         # the database.  (We don't consider this an LRU hit, though.)
         if (classname, nodeid) in self.cache: