diff --git a/ext/recover/dbdata.c b/ext/recover/dbdata.c index 225f1a9b3f..e755420ff1 100644 --- a/ext/recover/dbdata.c +++ b/ext/recover/dbdata.c @@ -511,11 +511,39 @@ static void dbdataValue( switch( enc ){ #ifndef SQLITE_OMIT_UTF16 case SQLITE_UTF16BE: - sqlite3_result_text16be(pCtx, (void*)pData, n, SQLITE_TRANSIENT); - break; - case SQLITE_UTF16LE: - sqlite3_result_text16le(pCtx, (void*)pData, n, SQLITE_TRANSIENT); + case SQLITE_UTF16LE: { + /* sqlite3_result_text16be/le() remove a leading BOM. If + ** the stored text begins with U+FEFF or U+FFFE, prepend + ** a BOM in the database byte order, so that only the + ** added BOM is removed. */ + u8 *aBuf = 0; + if( n>=2 && ((pData[0]==0xFE && pData[1]==0xFF) + || (pData[0]==0xFF && pData[1]==0xFE)) + ){ + aBuf = (u8*)sqlite3_malloc64(n+2); + if( aBuf==0 ){ + sqlite3_result_error_nomem(pCtx); + break; + } + aBuf[0] = (enc==SQLITE_UTF16BE) ? 0xFE : 0xFF; + aBuf[1] = (enc==SQLITE_UTF16BE) ? 0xFF : 0xFE; + memcpy(&aBuf[2], pData, n); + } + if( enc==SQLITE_UTF16BE ){ + if( aBuf ){ + sqlite3_result_text16be(pCtx, aBuf, n+2, sqlite3_free); + }else{ + sqlite3_result_text16be(pCtx, pData, n, SQLITE_TRANSIENT); + } + }else{ + if( aBuf ){ + sqlite3_result_text16le(pCtx, aBuf, n+2, sqlite3_free); + }else{ + sqlite3_result_text16le(pCtx, pData, n, SQLITE_TRANSIENT); + } + } break; + } #endif default: sqlite3_result_text(pCtx, (char*)pData, n, SQLITE_TRANSIENT); diff --git a/ext/recover/recover1.test b/ext/recover/recover1.test index 11a43780d9..a2fc7e4ed8 100644 --- a/ext/recover/recover1.test +++ b/ext/recover/recover1.test @@ -371,6 +371,60 @@ do_execsql_test 19.0 { do_recover_test 19.1 +#------------------------------------------------------------------------- +# Bug 2026-10-18T20:41:57Z +# +# In a UTF-16 database, TEXT values that begin with U+FEFF or U+FFFE +# must be recovered unchanged. The leading character is part of the +# value, not a byte-order mark. Values 4 and 5 are large enough to +# spill onto overflow pages. +# +foreach enc {utf16le utf16be} { + reset_db + do_execsql_test 20.$enc.0 " + PRAGMA encoding = '$enc'; + CREATE TABLE t1(id INTEGER PRIMARY KEY, a TEXT); + INSERT INTO t1 VALUES + (1, char(0xfeff)||'abc'), + (2, char(0xfffe)||'xyz'), + (3, 'plain'), + (4, char(0xfeff)||printf('%.*c', 5000, 'a')), + (5, char(0xfffe)||printf('%.*c', 5000, 'b')); + " + + set sql { + SELECT id, length(a), unicode(a), hex(substr(a,1,3)), hex(substr(a,-2)) + FROM t1 ORDER BY id + } + set expect [db eval $sql] + + # Recover directly into a new database. + do_test 20.$enc.1 { + forcedelete test.db2 + set R [sqlite3_recover_init db main test.db2] + $R run + $R finish + sqlite3 db2 test.db2 + set res [db2 eval $sql] + db2 close + set res + } $expect + + # Recover as SQL text (the path used by the CLI ".recover" command). + do_test 20.$enc.2 { + forcedelete test.db2 + set ::sqlhook [list] + set R [sqlite3_recover_init_sql db main my_sql_hook] + $R run + $R finish + sqlite3 db2 test.db2 + db2 eval [join $::sqlhook ";"] + set res [db2 eval $sql] + db2 close + set res + } $expect +} + finish_test diff --git a/manifest b/manifest index a8fdf7bf8e..457ff6c5a0 100644 --- a/manifest +++ b/manifest @@ -1,5 +1,5 @@ -C Fix\sthe\ssqlite3_bind_value()\sinterface\sso\sthat\sit\sdoes\snot\sremove\sa\nbyte-order\smark\son\sthe\svalue\sbeing\sbound. -D 2026-10-23T15:30:27.214 +C The\s.recover\scommand\sshould\snot\sremove\sBOMs\sthat\sare\salready\sinside\sthe\ndatabase.\n[bugs:/info/2026-10-18T20:41:57Z|Bug\s2026-10-18T20:41:57Z] +D 2026-10-23T15:52:17.165 F .fossil-settings/binary-glob 61195414528fb3ea9693577e1980230d78a1f8b0a54c78cf1b9b24d0a409ed6a x F .fossil-settings/empty-dirs dbb81e8fc0401ac46a1491ab34a7f2c7c0452f2f06b54ebb845d024ca8283ef1 F .fossil-settings/ignore-glob 35175cdfcf539b2318cb04a9901442804be81cd677d8b889fcc9149c21f239ea @@ -478,8 +478,8 @@ F ext/rbu/rbuvacuum4.test ffccd22f67e2d0b380d2889685742159dfe0d19a3880ca3d2d1d69 F ext/rbu/sqlite3rbu.c 29be4d2e7a7fbc8124545d2833570bdefef8724239c6cdd81a8d1e6dd07889f2 F ext/rbu/sqlite3rbu.h e3a5bf21e09ca93ce4e8740e00d6a853e90a697968ec0ea98f40826938bdb68e F ext/rbu/test_rbu.c 8b6e64e486c28c41ef29f6f4ea6be7b3091958987812784904f5e903f6b56418 -F ext/recover/dbdata.c 10d3c56968a9af6853722a47280805ad1564714d79ea45ac6f7da14bb57fd137 -F ext/recover/recover1.test e16d78e94183562abff569967b18b7c77451d7044365516cd0fe14713a284851 +F ext/recover/dbdata.c d49c08a6aad98a0da9b31d80826878cd77350b7ef380755adf8b8565c67755f0 +F ext/recover/recover1.test 876a31d02568c62c27186d560f5ffb23beaf9c268455fbe035627e23892aee4f F ext/recover/recover_common.tcl a61306c1eb45c0c3fc45652c35b2d4ec19729e340bdf65a272ce4c229cefd85a F ext/recover/recoverbuild.test c74170e0f7b02456af41838afeb5353fdb985a48cc2331d661bbabbca7c6b8e3 F ext/recover/recoverclobber.test 3ba6c0c373c5c63d17e82eced64c05c57ccaf26c1abe1ca7141334022a79f32e @@ -999,7 +999,7 @@ F test/bigmmap.test 6021e205487347c6d7e5a541aa472a4b8efc4e9f4a3799a823b61a8e6616 F test/bigrow.test 3e6b189dea465a78306eac67fc6754c1cf90eb1694a25c88c338900c216c1e6b F test/bigsort.test 997e172009905873c06426145e4b3794c7dfe2d563724cb2fd39d45f319cf3d2 F test/bind.test 1e136709b306f7ed3192d349c2930d89df6ab621654ad6f1a72381d3fe76f483 -F test/bind2.test 918bc35135f4141809ead7585909cde57d44db90a7a62aef540127148f91aab7 +F test/bind2.test e7914a9c670ff230d4fe6700d08becf4aff203ab2a3b48e4df6481e6c3526b77 F test/bindxfer.test efecd12c580c14df5f4ad3b3e83c667744a4f7e0 F test/bitvec.test 75894a880520164d73b1305c1c3f96882615e142 F test/blob.test e7ac6c7d3a985cc4678c64f325292529a69ae252 @@ -2297,11 +2297,8 @@ F tool/warnings-clang.sh bbf6a1e685e534c92ec2bfba5b1745f34fb6f0bc2a362850723a9ee F tool/warnings.sh a554d13f6e5cf3760f041b87939e3d616ec6961859c3245e8ef701d1eafc2ca2 F tool/win/sqlite.vsix deb315d026cc8400325c5863eef847784a219a2f F tool/winmain.c 00c8fb88e365c9017db14c73d3c78af62194d9644feaf60e220ab0f411f3604c -P 8a0296aa4db20cdcbbfaa0e4c61b5fe339b0514cf73832253ff0abd1ce59d31c -R 1dd3a5c88fdf1dc284f23dd932cd9767 -T *branch * bom-handling -T *sym-bom-handling * -T -sym-trunk * +P 56e268312d038608b5fb2a8ed24002f478ad5a22eb0214b1a064f2926aada66c +R 2e2481a1862ff08eb99bf69c8be110d9 U drh -Z e0a3b639a8ffa1ff10e4b189ad45981e +Z cc48029ffcfda22369666c438d0c5d3d # Remove this line to create a well-formed Fossil manifest. diff --git a/manifest.uuid b/manifest.uuid index f20994ff69..0b49f2dacb 100644 --- a/manifest.uuid +++ b/manifest.uuid @@ -1 +1 @@ -56e268312d038608b5fb2a8ed24002f478ad5a22eb0214b1a064f2926aada66c +69f4d423fa5538a555f862da3a1896e2fa3009dc650286042cb0beabe4dd2114 diff --git a/test/bind2.test b/test/bind2.test index 1ebf35c2e0..c9fed03717 100644 --- a/test/bind2.test +++ b/test/bind2.test @@ -36,6 +36,50 @@ do_test 1.1 { } {42.0} sqlite3_finalize $stmt + +# Bug 2026-10-18T20:41:57Z +# +# In a UTF-16 database, using sqlite3_bind_value() to copy a TEXT value +# that begins with U+FEFF or U+FFFE should copy the value exactly. The +# leading character is not a byte-order mark. +# +foreach enc {utf16le utf16be} { + forcedelete test.db3 + sqlite3 db3 test.db3 + do_execsql_test -db db3 2.$enc.0 " + PRAGMA encoding = '$enc'; + CREATE TABLE t2(id INTEGER PRIMARY KEY, a TEXT); + INSERT INTO t2 VALUES + (1, char(0xfeff)||'abc'), + (2, char(0xfffe)||'xyz'), + (3, 'plain'), + (4, char(0xfeff)||printf('%.*c', 5000, 'a')), + (5, char(0xfffe)||printf('%.*c', 5000, 'b')); + " + foreach {id res} { + 1 {4 65279 1} + 2 {4 65534 1} + 3 {5 112 1} + 4 {5001 65279 1} + 5 {5001 65534 1} + } { + do_test 2.$enc.$id { + set stmt [sqlite3_prepare db3 { + SELECT length(?1), unicode(?1), ?1=(SELECT a FROM t2 WHERE id=?2) + } -1 tail] + sqlite3_bind_int $stmt 2 $id + sqlite3_bind_value_from_select $stmt 1 "SELECT a FROM t2 WHERE id=$id" + sqlite3_step $stmt + set r [list [sqlite3_column_int $stmt 0] \ + [sqlite3_column_int $stmt 1] \ + [sqlite3_column_int $stmt 2]] + sqlite3_finalize $stmt + set r + } $res + } + db3 close +} + ifcapable !preupdate { finish_test return