Stefan (and Peter), The last thing I did last week was trying to isolate the cause of the bug from Hennings large example. I recreated the outerjoin in MIL and this gave exactly the same error. I add the script as an attachment, it may be easyer to find the bug from a small example, JanF. On Monday 23 April 2007 09:38, Stefan Manegold wrote:
Hi Henning,
thanks for your test --- according to Peter, the respective changes only tirggers the problem/bug (elsewhere), but is not the origin of the problem.bug itself, cf., http://sourceforge.net/tracker/index.php?func=detail&aid=1704972&group_id=5 6967&atid=482468
I suggested the test to find out whether the two problems (yours and the one mentioned in http://sourceforge.net/tracker/index.php?func=detail&aid=1704972&group_id=5 6967&atid=482468) are related --- apparently they are ...
Further, I don't have any idea what the actual problem is --- since we now have a way to reprocude it "easily" by running test pathfinder/benchmarks/X007/Tests/q19.xq, we'll have to single-step through the execution in gdb and try to locate the bug --- I *might* find some time for this towards the end of this week ...
Stefan
On Sun, Apr 22, 2007 at 03:49:42PM +0200, Henning Rode wrote:
hej stefan,
if i run the query with explicit call of the fn:exists() function, as you suggested, it works indeed fine.
could you please check, whether undoing Peter's change of MonetDB/src/gdk/gdk_relop.mx 1.29 -> 1.130 does solve your problem?
same, if i redo this change. the problem is gone then. since you suggested this test, i hope, you have an idea now, what goes wrong here. i don't think, i can repair it myself.
best -henning