Fix reading of most-negative integer value nodes

The main parser checks whether a literal fits into an int when
deciding whether it should be put into an Integer or Float node.  The
parser processes integer literals without signs.  So a most-negative
integer literal will not fit into Integer and will end up as a Float
node.

The node tokenizer did this differently.  It included the sign when
checking whether the literal fit into int.  So a most-negative integer
would indeed fit that way and end up as an Integer node.

In order to preserve the node structure correctly, we need the node
tokenizer to also analyze integer literals without sign.

There are a number of test cases in the regression tests that have a
most-negative integer argument of some utility statement, so this
issue is easily reproduced under WRITE_READ_PARSE_PLAN_TREES.

Reviewed-by: Tom Lane <[email protected]>
Discussion: https://www.postgresql.org/message-id/flat/4159834.1657405226@sss.pgh.pa.us

diff --git a/src/backend/nodes/read.c b/src/backend/nodes/read.c
index 4a54996b63f52f6772263f22cbe44ed133affde1..a9cb81b1290d9b14b901ef0e0bd91cef4c119728 100644 (file)
--- a/src/backend/nodes/read.c
+++ b/src/backend/nodes/read.c
@@ -267,7 +267,7 @@ nodeTokenType(const char *token, int length)
        char       *endptr;
 
        errno = 0;
-       (void) strtoint(token, &endptr, 10);
+       (void) strtoint(numptr, &endptr, 10);
        if (endptr != token + length || errno == ERANGE)
            return T_Float;
        return T_Integer;