Add //third_party/sqlite to dirs_to_snapshot, remove net_sql.patch

third_party/sqlite/src/test/fts3defer3.test

Index: third_party/sqlite/src/test/fts3defer3.test
diff --git a/third_party/sqlite/src/test/fts3defer3.test b/third_party/sqlite/src/test/fts3defer3.test
new file mode 100644
index 0000000000000000000000000000000000000000..a795e92bcf1ab8fcde102113258920c289a9e10c
--- /dev/null
+++ b/third_party/sqlite/src/test/fts3defer3.test
@@ -0,0 +1,84 @@
+# 2010 October 23
+#
+# The author disclaims copyright to this source code.  In place of
+# a legal notice, here is a blessing:
+#
+#    May you do good and not evil.
+#    May you find forgiveness for yourself and forgive others.
+#    May you share freely, never taking more than you give.
+#
+#***********************************************************************
+#
+# This file contains a very simple test to show that the deferred tokens
+# optimization is doing something.
+#
+
+set testdir [file dirname $argv0]
+source $testdir/tester.tcl
+source $testdir/malloc_common.tcl
+ifcapable !fts3||!fts4_deferred {
+  finish_test 
+  return
+}
+set testprefix fts3defer3
+
+set nDoclist 3204
+set nDoc     800
+
+# Set up a database that contains 800 rows. Each row contains the document
+# "b b", except for the row with docid=200, which contains "a b". Hence
+# token "b" is extremely common and token "a" is not.
+#
+do_test 1.1 {
+  execsql {
+    CREATE VIRTUAL TABLE t1 USING fts4;
+      BEGIN;
+  }
+  for {set i 1} {$i <= $nDoc} {incr i} {
+    set document "b b"
+    if {$i==200} { set document "a b" }
+    execsql { INSERT INTO t1 (docid, content) VALUES($i, $document) }
+  }
+  execsql COMMIT
+} {}
+
+# Check that the db contains two doclists. A small one for "a" and a 
+# larger one for "b".
+#
+do_execsql_test 1.2 {
+  SELECT blockid, length(block) FROM t1_segments;
+} [list 1 8 2 $nDoclist]
+
+# Query for 'a b'. Although this test doesn't prove so, token "b" will 
+# be deferred because of the very large associated doclist.
+#
+do_execsql_test 1.3 {
+  SELECT docid, content FROM t1 WHERE t1 MATCH 'a b';
+} {200 {a b}}
+
+# Zero out the doclist for token "b" within the database file. Now the 
+# only queries that use token "b" that will work are those that defer
+# it. Any query that tries to use the doclist belonging to token "b"
+# will fail.
+#
+do_test 1.4 {
+  set fd [db incrblob t1_segments block 2]
+  puts -nonewline $fd [string repeat "\00" $nDoclist]
+  close $fd
+} {}
+
+# The first two queries succeed, as they defer token "b". The last one
+# fails, as it tries to load the corrupt doclist.
+#
+do_execsql_test 1.5 {
+  SELECT docid, content FROM t1 WHERE t1 MATCH 'a b';
+} {200 {a b}}
+do_execsql_test 1.6 {
+  SELECT count(*) FROM t1 WHERE t1 MATCH 'a b';
+} {1}
+do_catchsql_test 1.7 {
+  SELECT count(*) FROM t1 WHERE t1 MATCH 'b';
+} {1 {database disk image is malformed}}
+
+
+finish_test