session-1:
vac=> begin;
BEGIN
vac=> lock table t in row share mode;
LOCK TABLE
vac=>
session-2:
vac=> begin;
BEGIN
vac=> lock table t in exclusive mode;
-- waiting
session-3:
vac=> begin;
BEGIN
vac=> lock table t in row share mode;
LOCK TABLE
vac=>
???
First, I agree to give some transaction extra priority
while locking table A but _only_ if transaction already
has any kind of lock on _this_ table. I object to do this
if transaction has only locks on _other_ tables - this
will result in starving.
Second, session-3 didn't hold any kind of locks but
MyProc->lockQueue wasn't empty! Why?
And last:
try to press Ctrl-D in session-3 and you'll get
NOTICE: Message from PostgreSQL backend: The Postmaster has informed me that some other backend died
abnormally and possibly corrupted shared memory.
in session-2. Core made for session-3:
#5 0x8114fea in ExceptionalCondition ( conditionName=0x814f9d8 "!((((unsigned long)queue) > ShmemBase))",
exceptionP=0x816490c,detail=0x0, fileName=0x814f9cd "shmqueue.c", lineNumber=248) at assert.c:74
#6 0x80defc3 in SHMQueueEmpty (queue=0x48) at shmqueue.c:248
#7 0x80e2be4 in LockResolveConflicts (lockmethod=1, lock=0x30252d68, lockmode=6, xid=34309, xidentP=0x0) at
lock.c:832
#8 0x80e4690 in ProcLockWakeup (queue=0x30252d7c, lockmethod=1, lock=0x30252d68) at proc.c:683
#9 0x80e3ade in LockReleaseAll (lockmethod=1, lockQueue=0x30252ce8) at lock.c:1456
#10 0x80e438d in ProcKill (exitStatus=0, pid=24636) at proc.c:410
#11 0x80dd92c in shmem_exit (code=0) at ipc.c:193
#12 0x80dd861 in proc_exit (code=0) at ipc.c:139
#13 0x80e8a8c in PostgresMain (argc=11, argv=0xefbfccfc, real_argc=6, real_argv=0xefbfd774) at postgres.c:1672
Vadim
