Do not save the state of an fts5 merge operation mid-way through an input

term, even if no previous entry for that term has caused any output. Doing so
may corrupt the FTS index.

FossilOrigin-Name: 9a2de4f05fabf7e725666fb8824ae1adad56834d7ae4f2e118375bb6f9c0ae77

diff --git a/ext/fts5/fts5_index.c b/ext/fts5/fts5_index.c
index 3d6616bd91d5c206d5c62281217629365c4d1f72..6dc9f8470e107932f046fb5779dbf1ec81497930 100644 (file)
--- a/ext/fts5/fts5_index.c
+++ b/ext/fts5/fts5_index.c
@@ -4192,6 +4192,7 @@ static void fts5IndexMergeLevel(
   int bOldest;                    /* True if the output segment is the oldest */
   int eDetail = p->pConfig->eDetail;
   const int flags = FTS5INDEX_QUERY_NOOUTPUT;
+  int bTermWritten = 0;           /* True if current term already output */
 
   assert( iLvl<pStruct->nLevel );
   assert( pLvl->nMerge<=pLvl->nSeg );
@@ -4245,18 +4246,22 @@ static void fts5IndexMergeLevel(
     int nTerm;
     const u8 *pTerm;
 
-    /* Check for key annihilation. */
-    if( pSegIter->nPos==0 && (bOldest || pSegIter->bDel==0) ) continue;
-
     pTerm = fts5MultiIterTerm(pIter, &nTerm);
     if( nTerm!=term.n || memcmp(pTerm, term.p, nTerm) ){
       if( pnRem && writer.nLeafWritten>nRem ){
         break;
       }
+      fts5BufferSet(&p->rc, &term, nTerm, pTerm);
+      bTermWritten =0;
+    }
 
+    /* Check for key annihilation. */
+    if( pSegIter->nPos==0 && (bOldest || pSegIter->bDel==0) ) continue;
+
+    if( p->rc==SQLITE_OK && bTermWritten==0 ){
       /* This is a new term. Append a term to the output segment. */
       fts5WriteAppendTerm(p, &writer, nTerm, pTerm);
-      fts5BufferSet(&p->rc, &term, nTerm, pTerm);
+      bTermWritten = 1;
     }
 
     /* Append the rowid to the output */

diff --git a/ext/fts5/test/fts5delete.test b/ext/fts5/test/fts5delete.test
new file mode 100644 (file)
index 0000000..15d62c5
--- /dev/null
+++ b/ext/fts5/test/fts5delete.test
@@ -0,0 +1,54 @@
+# 2017 May 12
+#
+# 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 implements regression tests for SQLite library.  The
+# focus of this script is testing the FTS5 module.
+#
+
+source [file join [file dirname [info script]] fts5_common.tcl]
+set testprefix fts5delete
+
+# If SQLITE_ENABLE_FTS5 is not defined, omit this file.
+ifcapable !fts5 {
+  finish_test
+  return
+}
+fts5_aux_test_functions db
+
+do_execsql_test 1.0 {
+  CREATE VIRTUAL TABLE t1 USING fts5(x);
+  WITH s(i) AS (
+    SELECT 1 UNION ALL SELECT i+1 FROM s WHERE i<5000
+  )
+  INSERT INTO t1(rowid, x) SELECT i, (i/2)*2 FROM s;
+}
+
+do_test 1.1 {
+  execsql BEGIN
+  for {set i 1} {$i<=5000} {incr i} {
+    if {$i % 2} {
+      execsql { INSERT INTO t1 VALUES($i) }
+    } else {
+      execsql { DELETE FROM t1 WHERE rowid = $i }
+    }
+  }
+  execsql COMMIT
+} {}
+
+do_test 1.2 {
+  execsql { INSERT INTO t1(t1, rank) VALUES('usermerge', 2); }
+  for {set i 0} {$i < 5} {incr i} {
+    execsql { INSERT INTO t1(t1, rank) VALUES('merge', 1) }
+    execsql { INSERT INTO t1(t1) VALUES('integrity-check') }
+  }
+} {}
+
+finish_test
+

diff --git a/manifest b/manifest
index de491ca0b8ac925dd4bc25e7f4855f5cbc8136a7..d5d926e20503a0c7bb0ef5bdacd7864848632229 100644 (file)
--- a/manifest
+++ b/manifest
@@ -1,5 +1,5 @@
-C New\srequirements\smarks\sand\sdocumentation\sfor\sthe\sauthorizer.
-D 2017-05-11T13:43:57.931
+C Do\snot\ssave\sthe\sstate\sof\san\sfts5\smerge\soperation\smid-way\sthrough\san\sinput\nterm,\seven\sif\sno\sprevious\sentry\sfor\sthat\sterm\shas\scaused\sany\soutput.\sDoing\sso\nmay\scorrupt\sthe\sFTS\sindex.
+D 2017-05-11T18:14:06.985
 F Makefile.in 1cc758ce3374a32425e4d130c2fe7b026b20de5b8843243de75f087c0a2661fb
 F Makefile.linux-gcc 7bc79876b875010e8c8f9502eb935ca92aa3c434
 F Makefile.msc 6a8c838220f7c00820e1fc0ac1bccaaa8e5676067e1dbfa1bafa7a4ffecf8ae6
@@ -105,7 +105,7 @@ F ext/fts5/fts5_buffer.c 4c1502d4c956cd092c89ce4480867f9d8bf325cd
 F ext/fts5/fts5_config.c 5af9c360e99669d29f06492c370892394aba0857
 F ext/fts5/fts5_expr.c f2825f714d91bbe62ab5820aee9ad12e0c94205b2a01725eaa9072415ae9ff1c
 F ext/fts5/fts5_hash.c 880998e596b60f078348d48732ca4ad9a90caad2
-F ext/fts5/fts5_index.c dc25123df20c60492857de491a194dab4b46ace217b8483bda305d357bf6431d
+F ext/fts5/fts5_index.c 9ce10106f42f8b84278a8ea859940224e2af5f0cc882f909364469f6f52769cb
 F ext/fts5/fts5_main.c 1ba0e7806886c1bc16e20d0dde1c2b535d1aeb98cbbb937c4c3e064af5ac6f03
 F ext/fts5/fts5_storage.c 7750986004f3f0c94619a85ecb5dd6cbef53e5e3853488e8a906c269d4d11db6
 F ext/fts5/fts5_tcl.c 4a901f00c8553740dba63511603f5527d741c26a
@@ -144,6 +144,7 @@ F ext/fts5/test/fts5content.test 9a952c95518a14182dc3b59e3c8fa71cda82a4e1
 F ext/fts5/test/fts5corrupt.test c2ad090192708150d50d961278df10ae7a4b8b62
 F ext/fts5/test/fts5corrupt2.test 128eb6e2d26b09f4da339e581f424b3321e0fdaa
 F ext/fts5/test/fts5corrupt3.test f77f65e386231daf62902466b40ff998b2c8ce4f
+F ext/fts5/test/fts5delete.test 0585395660889090a6a04b0cac0e70b16801da44ee321534d6169da49af8167f
 F ext/fts5/test/fts5detail.test ef5c690535a797413acaf5ad9b8ab5d49972df69
 F ext/fts5/test/fts5determin.test 10648edb75ef1e196b10978fd21a9be0c31e09c3
 F ext/fts5/test/fts5dlidx.test 007e9390c94638760797dbec2990c97c3fa08dfe
@@ -1579,7 +1580,7 @@ F vsixtest/vsixtest.tcl 6a9a6ab600c25a91a7acc6293828957a386a8a93
 F vsixtest/vsixtest.vcxproj.data 2ed517e100c66dc455b492e1a33350c1b20fbcdc
 F vsixtest/vsixtest.vcxproj.filters 37e51ffedcdb064aad6ff33b6148725226cd608e
 F vsixtest/vsixtest_TemporaryKey.pfx e5b1b036facdb453873e7084e1cae9102ccc67a0
-P 47629b1911e52445aad8ea969137bddf0019c55b4a4f0de8e77decb6a434c8a2
-R 0c6451e367b2cefc9e69fa0621b4d931
-U drh
-Z 18c2f8082a296f2f15ab41c2498f7cb3
+P 3980ea0911b3ad3f86d7a7bdc6503f233315c274f473e18831e13eda2c238eeb
+R e8ce1902ba436a257e96b01caa3cadb9
+U dan
+Z c1b8197d958a390da1753ec3ab7dee72

diff --git a/manifest.uuid b/manifest.uuid
index 4e27390c060afca32da75bc46c40b8659255bd40..8a1a73a65b1214a1aec21ab1321e6cd16b0cf708 100644 (file)
--- a/manifest.uuid
+++ b/manifest.uuid
@@ -1 +1 @@
-3980ea0911b3ad3f86d7a7bdc6503f233315c274f473e18831e13eda2c238eeb
\ No newline at end of file
+9a2de4f05fabf7e725666fb8824ae1adad56834d7ae4f2e118375bb6f9c0ae77
\ No newline at end of file