Hi,
On Debian and amd64 architecture SQLite3 has a severe problem. If I just
start it, I can create a simple table like "create table a(b int);".
However when I set "export MALLOC_CHECK_=1" then the mentioned table
creation reports:
*** glibc detected *** sqlite3: free(): invalid pointer: 0x00007fba85c30560 ***
If MALLOC_CHECK_ set to 2 then it just aborts with a zero length
database and a database-journal being 512 bytes long.
gdb reports the following related backtrace:
#2 0x00007ffff75ddae0 in malloc_printerr (action=2, str=0x7ffff76b186a
"free(): invalid pointer", ptr=0x2399) at malloc.c:6288
#3 0x00007ffff7b4d070 in sqlite3_free (p=0x555555791e10) at sqlite3.c:19059
#4 sqlite3_free (p=0x555555791e10) at sqlite3.c:19051
#5 0x00007ffff7bb1312 in sqlite3VdbeExec (p=0x555555793e10) at sqlite3.c:69790
#6 0x00007ffff7b9a731 in sqlite3Step (p=0x555555793e10) at sqlite3.c:63163
#7 sqlite3_step (pStmt=0x555555793e10) at sqlite3.c:63236
I tried to set a breakpoint on sqlite3VdbeExec and see what's going on.
Then p is (Vdbe *) 0x5555557950c0 and its db part (print *p->db):
{pVfs = 0x7ffff7dda2a0, pVdbe = 0x5555557950c0, pDfltColl = 0x5555557772b0,
mutex = 0x0, aDb = 0x555555780c80, nDb = 2,
flags = 1744833024, lastRowid = 0, openFlags = 6, errCode = 0, errMask = 255,
autoCommit = 1 '\001', temp_store = 0 '\000',
mallocFailed = 0 '\000', dfltLockMode = 0 '\000', nextAutovac = -1 'ΓΏ',
suppressErr = 0 '\000', vtabOnConflict = 0 '\000',
isTransactionSavepoint = 0 '\000', nextPagesize = 0, magic = 2687084183,
nChange = 0, nTotalChange = 0, aLimit = {1000000000, 1000000000,
2000, 1000, 500, 25000, 127, 10, 50000, 999, 1000}, init = {newTnum = 1,
iDb = 0 '\000', busy = 1 '\001', orphanTrigger = 0 '\000'},
activeVdbeCnt = 1, writeVdbeCnt = 0, vdbeExecCnt = 1, nExtension = 0,
aExtension = 0x0, xTrace = 0, pTraceArg = 0x0, xProfile = 0,
pProfileArg = 0x0, pCommitArg = 0x0, xCommitCallback = 0, pRollbackArg = 0x0,
xRollbackCallback = 0, pUpdateArg = 0x0,
xUpdateCallback = 0, xWalCallback = 0x7ffff7bbbc70 <sqlite3WalDefaultHook>,
pWalArg = 0x3e8, xCollNeeded = 0, xCollNeeded16 = 0,
pCollNeededArg = 0x0, pErr = 0x555555777380, zErrMsg = 0x0, zErrMsg16 = 0x0,
u1 = {isInterrupted = 0, notUsed1 = 0}, lookaside = {
sz = 128, bEnabled = 0 '\000', bMalloced = 1 '\001', nOut = 0, mxOut = 0,
anStat = {0, 0, 0}, pFree = 0x555555791ac0,
pStart = 0x555555782140, pEnd = 0x555555791b40}, xAuth = 0, pAuthArg = 0x0,
xProgress = 0, pProgressArg = 0x0, nProgressOps = 0,
nVTrans = 0, aModule = {htsize = 0, count = 5, first = 0x555555782100, ht =
0x0}, pVtabCtx = 0x0, aVTrans = 0x0, pDisconnect = 0x0,
aFunc = {a = {0x0, 0x5555557819f0, 0x0, 0x555555781d50, 0x555555781e30, 0x0,
0x0, 0x555555781c90, 0x555555781f60, 0x555555781fd0, 0x0,
0x555555791b50, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0,
0x5555557810b0}}, aCollSeq = {htsize = 0, count = 3,
first = 0x555555780df0, ht = 0x0}, busyHandler = {xFunc = 0, pArg = 0x0,
nBusy = 0}, aDbStatic = {{zName = 0x7ffff7bbdafe "main",
pBt = 0x555555780e30, inTrans = 0 '\000', safety_level = 3 '\003',
pSchema = 0x555555780f90}, {zName = 0x7ffff7bbd0ca "temp",
pBt = 0x0, inTrans = 0 '\000', safety_level = 1 '\001', pSchema =
0x555555781020}}, pSavepoint = 0x0, busyTimeout = 0,
nSavepoint = 0, nStatement = 0, nDeferredCons = 0, pnBytesFreed = 0x0,
pBlockingConnection = 0x0, pUnlockConnection = 0x0,
pUnlockArg = 0x0, xUnlockNotify = 0, pNextBlocked = 0x0}
After four continue it aborts with the backtrace mentioned on top.
The compilation options (CFLAGS) were:
-O2 -fno-strict-aliasing
-DSQLITE_SECURE_DELETE
-DSQLITE_ENABLE_COLUMN_METADATA
-DSQLITE_ENABLE_FTS3 -DSQLITE_ENABLE_RTREE=1
-DSQLITE_SOUNDEX=1
-DSQLITE_ENABLE_UNLOCK_NOTIFY
-DSQLITE_OMIT_LOOKASIDE=1
-DSQLITE_ENABLE_UPDATE_DELETE_LIMIT=1
Removing -DSQLITE_OMIT_LOOKASIDE=1 helps a bit, the basic table creation
works. It still doesn't pass the SQLite3 tests during compilation of
bogofilter 1.2.2+dfsg1 . The Debian bugreport #665363 [1] states that
3.7.9 was working and the problem first noted in 3.7.10 .
What should I do to help it being fixed? Would you need *p->db _right
before_ the abort?
Regards,
Laszlo/GCS
[1] http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=665363
_______________________________________________
sqlite-users mailing list
sqlite-users@sqlite.org
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
