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 f63fc68152..3610ecac87 100644 --- a/manifest +++ b/manifest @@ -1,5 +1,5 @@ -C FTS5:\sFix\sfalse-positive\scorrupt\sFTS5\sindex\sreport\son\san\sINSERT\sinto\na\scontentless_unindexed\stable.\n[bugs:/info/2026-10-21T02:49:37Z|Bug\s2026-10-21T02:49:37Z] -D 2026-10-23T13:41:31.649 +C Merge\sBOM\shandling\sfixes. +D 2026-10-23T15:56:07.723 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 @@ -809,10 +809,10 @@ F src/resolve.c 658d5d8f67b7565cbed56394d0fdbb124561a92c7455253186189db5e090e661 F src/rowset.c 8432130e6c344b3401a8874c3cb49fefe6873fec593294de077afea2dce5ec97 F src/select.c 114c5874d9c61bb4175831067b6f7c2542da72c9e5fba10ff944c9ebded8bcea F src/shell.c.in 808c0bd4a593a3552777c059e90c9c47e459f507fa25d2afd298edf09d3140f5 -F src/sqlite.h.in 59fcb8ca48dd0477770c32b653ac67f910a4eefadfe0c98868b7a9cbcff95019 +F src/sqlite.h.in 5588081b98e60a771c6ee3512a9f5837ee3268130e524d5dc956601eafc737ae F src/sqlite3.rc 015537e6ac1eec6c7050e17b616c2ffe6f70fca241835a84a4f0d5937383c479 F src/sqlite3ext.h 5e113c4c0442f2f1b3345d44cbe4b82fad45f2fc0c35f0ed4eef27fb8f1cb394 -F src/sqliteInt.h 97293a49107ea79c7ebe72f9a5ffc262a3ecb8bf59ba0002c1f4753669bd2e3d +F src/sqliteInt.h 5a4faf7d7b6103df8797240cad67f9f63b77c8043487f23c6eac7dd62a0667a0 F src/sqliteLimit.h a905209cea2a167391382f64ecd42f944863ae44eabd540ef24a70361182f5ae F src/status.c 7565d63a79aa2f326339a24a0461a60096d0bd2bce711fefb50b5c89335f3592 F src/table.c 0f141b58a16de7e2fbe81c308379e7279f4c6b50eb08efeec5892794a0ba30d1 @@ -874,13 +874,13 @@ F src/upsert.c dd9f0fcccbfb4f20e1026a21a7254ba3f2c08e9cfa92affaff5b5ec3b00ea549 F src/utf.c c5bec56770b71fe75e12f1f7b0a512032e1fd3d1f83c50c4ab4474ced20347a4 F src/util.c 5cda82abda6cc880f67749af782ebd79ac1fad75edf9ab8cdb568482a71e9cd4 F src/vacuum.c d3d35d8ae893d419ade5fa196d761a83bddcbb62137a1a157ae751ef38b26e82 -F src/vdbe.c 7bb9e161f05986446186513a5c9498455930069c2177b12e074a46fa006890c5 +F src/vdbe.c 3ec0ab7a9eff7e7ad32e43a5178c35030e1cb505e4866b707897e87fad9bb8ca F src/vdbe.h b6ffb28da04363fd5d439b8204118fd3f2e751abe4f8e3edb089155fb90d4607 F src/vdbeInt.h 581b112c6a763a5871ae1d2b948c5288e749d03c7e3829ebce2738402fe21074 -F src/vdbeapi.c 4b38eb061fe6a8e96055e74d9301f0f36e453855e650a3e5fdd25ad45c7d0d7e +F src/vdbeapi.c b1501e7184c5286cf2507e80bd437e8ea749997376bacaf199160c8563ef1d16 F src/vdbeaux.c 7e3975222531284ff14cb1be93b5751f6a3392a56d86d6329fcaa120471f56f3 F src/vdbeblob.c b3f0640db9642fbdc88bd6ebcc83d6009514cafc98f062f675f2c8d505d82692 -F src/vdbemem.c cf4000a8b355be2028b290516898ff99d611c0d33397ac6c4f6d77f84dcd39fb +F src/vdbemem.c 29f9bf1f900c85415182e7f97cfcee296742297b3f93113927f3767eeedd8df2 F src/vdbesort.c a72ad2a629df2901fde4aa588e38590d0aa27b7d5bc3b2a416e5f295de839cc1 F src/vdbetrace.c 49e689f751505839742f4a243a1a566e57d5c9eaf0d33bbaa26e2de3febf7b41 F src/vdbevtab.c fc46b9cbd759dc013f0b3724549cc0d71379183c667df3a5988f7e2f1bd485f3 @@ -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 @@ -2297,8 +2297,9 @@ 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 a985a224b90b6f325d171c0f16364101 +P 887264cd9a2ad95d808f910b4d0be04e8435382eb7640b4d4769b648b266d3a8 69f4d423fa5538a555f862da3a1896e2fa3009dc650286042cb0beabe4dd2114 +R 04459a9ef61bcf00c752bbf32de5526e +T +closed 69f4d423fa5538a555f862da3a1896e2fa3009dc650286042cb0beabe4dd2114 U drh -Z 94e6afe38d2ae8ef5ce16498ad41fb23 +Z 7083e87ae61fbcde4ffc08f5b9bcd8cb # Remove this line to create a well-formed Fossil manifest. diff --git a/manifest.uuid b/manifest.uuid index 3ae6bcc9f8..40f9e3e48f 100644 --- a/manifest.uuid +++ b/manifest.uuid @@ -1 +1 @@ -887264cd9a2ad95d808f910b4d0be04e8435382eb7640b4d4769b648b266d3a8 +5b0c5f63416a671ee15a2a2bae430d059ea9b664c0f2db41d033286d00d1a56b diff --git a/src/sqlite.h.in b/src/sqlite.h.in index 00fbe00f46..451e3203ff 100644 --- a/src/sqlite.h.in +++ b/src/sqlite.h.in @@ -5888,6 +5888,10 @@ int sqlite3_create_window_function( ** These constants define integer codes that represent the various ** text encodings supported by SQLite. ** +** If any value other than the ones defined here are passed into the +** "encoding" argument of an API function, then the results are +** unspecified and subject to change. +** **
** [[SQLITE_UTF8]]
SQLITE_UTF8
Text is encoding as UTF-8
** diff --git a/src/sqliteInt.h b/src/sqliteInt.h index 6c64114a36..1e3899d57d 100644 --- a/src/sqliteInt.h +++ b/src/sqliteInt.h @@ -1843,10 +1843,13 @@ struct sqlite3 { }; /* -** A macro to discover the encoding of a database. +** A macro to discover the encoding of a database or for managing +** encoding values. */ #define SCHEMA_ENC(db) ((db)->aDb[0].pSchema->enc) #define ENC(db) ((db)->enc) +#define ENC_MASK 0x1f /* Usable encoding bits */ +#define ENC_KEEPBOM 0x80 /* Do not remove BOM */ /* ** A u64 constant where the lower 32 bits are all zeros. Only the diff --git a/src/vdbe.c b/src/vdbe.c index 8a1f6687f0..c56bfb21e0 100644 --- a/src/vdbe.c +++ b/src/vdbe.c @@ -796,18 +796,14 @@ static SQLITE_NOINLINE int vdbeColumnFromOverflow( assert( t>=12 ); sqlite3RCStrRef(pBuf); if( t&1 ){ - /* Prevents the removal of a BOM in UTF16 text. Otherwise - ** it causes no harm as long as len>=0, and pDest->enc is set - ** to the correct encoding afterwards: ---- - ** (tag-20261016-1) `vvvvvvvvvvv */ - rc = sqlite3VdbeMemSetStr(pDest, pBuf, len, SQLITE_UTF8, + rc = sqlite3VdbeMemSetStr(pDest, pBuf, len, encoding|ENC_KEEPBOM, sqlite3RCStrUnref); pDest->flags |= MEM_Term; }else{ rc = sqlite3VdbeMemSetStr(pDest, pBuf, len, 0, sqlite3RCStrUnref); - } - pDest->enc = encoding; + pDest->enc = encoding; + } }else{ rc = sqlite3VdbeMemFromBtree(pC->uc.pCursor, iOffset, len, pDest); if( rc ) return rc; diff --git a/src/vdbeapi.c b/src/vdbeapi.c index abdbb342f4..bc00881417 100644 --- a/src/vdbeapi.c +++ b/src/vdbeapi.c @@ -1784,12 +1784,13 @@ static int bindText( const void *zData, /* Pointer to the data to be bound */ i64 nData, /* Number of bytes of data to be bound */ void (*xDel)(void*), /* Destructor for the data */ - u8 encoding /* Encoding for the data */ + u8 encoding /* Encoding for the data + optional ENC_KEEPBOM */ ){ Vdbe *p = (Vdbe *)pStmt; Mem *pVar; int rc; + assert( (encoding&ENC_KEEPBOM)==0 || (encoding&ENC_MASK)>SQLITE_UTF8 ); rc = vdbeUnbind(p, (u32)(i-1)); if( rc==SQLITE_OK ){ assert( p!=0 && p->aVar!=0 && i>0 && i<=p->nVar ); /* tag-20240917-01 */ @@ -1996,8 +1997,9 @@ int sqlite3_bind_value(sqlite3_stmt *pStmt, int i, const sqlite3_value *pValue){ break; } case SQLITE_TEXT: { + assert( pValue->enc!=0 ); rc = bindText(pStmt,i, pValue->z, pValue->n, SQLITE_TRANSIENT, - pValue->enc); + pValue->enc<=SQLITE_UTF8 ? pValue->enc : pValue->enc|ENC_KEEPBOM); break; } default: { diff --git a/src/vdbemem.c b/src/vdbemem.c index 258c773803..fdf56673e0 100644 --- a/src/vdbemem.c +++ b/src/vdbemem.c @@ -1307,11 +1307,13 @@ int sqlite3VdbeMemSetStr( i64 nByte = n; /* New value for pMem->n */ int iLimit; /* Maximum allowed string or blob size */ u16 flags; /* New value for pMem->flags */ + u8 xenc; /* Encoding without the ENC_KEEPBOM bit */ assert( pMem!=0 ); assert( pMem->db==0 || sqlite3_mutex_held(pMem->db->mutex) ); assert( !sqlite3VdbeMemIsRowSet(pMem) ); - assert( enc!=0 || n>=0 ); + xenc = enc & ENC_MASK; + assert( xenc!=0 || n>=0 ); /* If z is a NULL pointer, set pMem to contain an SQL NULL. */ if( !z ){ @@ -1325,16 +1327,16 @@ int sqlite3VdbeMemSetStr( iLimit = SQLITE_MAX_LENGTH; } if( nByte<0 ){ - assert( enc!=0 ); - if( enc==SQLITE_UTF8 ){ + assert( xenc!=0 ); + if( xenc==SQLITE_UTF8 ){ nByte = strlen(z); }else{ for(nByte=0; nByte<=iLimit && (z[nByte] | z[nByte+1]); nByte+=2){} } flags= MEM_Str|MEM_Term; - }else if( enc==0 ){ + }else if( xenc==0 ){ flags = MEM_Blob; - enc = SQLITE_UTF8; + xenc = SQLITE_UTF8; }else{ flags = MEM_Str; } @@ -1357,7 +1359,7 @@ int sqlite3VdbeMemSetStr( if( xDel==SQLITE_TRANSIENT ){ i64 nAlloc = nByte; if( flags&MEM_Term ){ - nAlloc += (enc==SQLITE_UTF8?1:2); + nAlloc += (xenc==SQLITE_UTF8?1:2); } testcase( nAlloc==0 ); testcase( nAlloc==31 ); @@ -1381,14 +1383,13 @@ int sqlite3VdbeMemSetStr( pMem->n = (int)(nByte & 0x7fffffff); pMem->flags = flags; - pMem->enc = enc; + pMem->enc = xenc; #ifndef SQLITE_OMIT_UTF16 - /* tag-20261016-1: To retain the BOM, use enc==SQLITE_UTF8, even - ** for UTF-16 text, and then set pMem->enc correctly after this - ** function returns. Note, however, that n must be non-negative - ** for this trick to work. */ - if( enc>SQLITE_UTF8 && sqlite3VdbeMemHandleBom(pMem) ){ + if( xenc>SQLITE_UTF8 + && (enc & ENC_KEEPBOM)==0 + && sqlite3VdbeMemHandleBom(pMem) + ){ return SQLITE_NOMEM_BKPT; } #endif 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