serialization of negative varints is broken

[espindola]

Running the following CQL on an empty database

CREATE KEYSPACE test_ks WITH REPLICATION = { 'class' : 'SimpleStrategy' , 'replication_factor' : 1 };                                                                                                                                      
use test_ks;                                                                                                                                                                                                                               
CREATE TABLE my_table (val varint PRIMARY KEY) with compression = {'sstable_compression': ''};                                                                                                                                             
INSERT INTO my_table (val) VALUES (-2147483648); 

In scylla produces the sstable

00 05                   key_length                                                                                                                                                                                                         
ff 80 00 00 00          key                                                                                                                                                                                                                
7f ff ff ff             local_deletion_time                                                                                                                                                                                                
80 00 00 00 00 00 00 00 marked_for_delete_at                                                                                                                                                                                               
24                      HAS_ALL_COLUMNS | HAS_TIMESTAMP                                                                                                                                                                                                          
02                      row_body_size
13                      prev_unfiltered_size
00                      delta_timestamp
01                      END_OF_PARTITION

With cassandra I get

00 04                   key_length                                                                                                                                                                                                         
80 00 00 00             key                                                                                                                                                                                                                
7f ff ff ff             local_deletion_time                                                                                                                                                                                                
80 00 00 00 00 00 00 00 marked_for_delete_at                                                                                                                                                                                               
24                      HAS_ALL_COLUMNS | HAS_TIMESTAMP                                                                                                                                                                                                          
02                      row_body_size
12                      prev_unfiltered_size
00                      delta_timestamp
01                      END_OF_PARTITION

Note how the scylla sstable has an extra 0xff at the start of the varint value.

When prepared statements are used, scylla writes the value produced by the driver, so the bug is hidden.

[espindola]

Should this be backported?

[avikivity]

Backported to 3.3, 3.2, 3.1.

[tzach]

reproducer by @espindola

#!/usr/bin/python

from cassandra.cluster import Cluster

cluster = Cluster()
session = cluster.connect()

session.execute("CREATE KEYSPACE test_ks WITH REPLICATION = { 'class' : 'SimpleStrategy' , 'replication_factor' : 1 };")
session.execute("use test_ks;")
session.execute("CREATE TABLE my_table (val varint PRIMARY KEY) with compression = {'sstable_compression': ''};")

p = session.prepare("INSERT INTO my_table (val) VALUES (?);")

session.execute(p.bind([-2147483648]))

session.execute("INSERT INTO my_table (val) VALUES (-2147483648);")

print(session.execute("SELECT * FROM my_table;"))