Hi Stefan, I deleted the entire 'prefix' directory and did a 'make install' for (buildtools,MonetDB,pathfinder) and now it works OK. So the old module files indeed caused the problem, Thanks, Jan. On Friday 13 October 2006 10:06, Stefan Manegold wrote:

Hi Jan,

[second attempt, now with correctly spelled email adresses]

[since other user might have/run into the same problem, I feel free to cc my reply to monetdb-users@lists.sourceforge.net.]

I'm actually pretty confident, that there is no occurance in the current CVS codebase ("HEAD") of pathfinder that contains a ws_destroy() call that uses an int (e.g., "int(ws)") instead of the ws BAT itself (i.e., "ws") as argument:

======== $ find pathfinder/ -type f | xargs grep --color ws_destroy pathfinder/runtime/Tests/procs.stable.out:[ "ws_destroy(BAT[void,BAT]) : void" ] pathfinder/runtime/Tests/xrpc.stable.out:[ "ws_destroy(BAT[void,BAT]) : void" ] pathfinder/runtime/pathfinder.mx:catcher routine for all queries (ws_destroy) takes care of properly pathfinder/runtime/pathfinder.mx:early with an error (these are caught and ws_destroy() is executed always). pathfinder/runtime/pathfinder.mx:PROC ws_destroy(BAT[void,bat] ws) : void pathfinder/runtime/pathfinder.mx: doc_undo.delete(doc_undo.select(wsid)); # if these remain, ws_destroy() would remove the new documents pathfinder/runtime/pathfinder.mx:bat. Only if everything succeeds, the entries are removed from there. Otherwise, in ws_destroy(), pathfinder/runtime/pathfinder.mx:running (this is triggered by the last leaving query in ws_destroy()). pathfinder/tests/Update/Tests/insert-2.stable.out:ws_destroy(ws); pathfinder/tests/Update/Tests/setattr-1.stable.out:ws_destroy(ws); pathfinder/tests/Update/Tests/symm-test-1.epilogue:ws_destroy(ws); pathfinder/tests/Update/Tests/symm-test-1.stable.out:ws_destroy(ws); pathfinder/tests/Update/Tests/update.epilogue:ws_destroy(ws); pathfinder/tests/Update/Tests/insert-1.stable.out:ws_destroy(ws); pathfinder/tests/Update/Tests/symm-test-2.stable.out:ws_destroy(ws); pathfinder/tests/Update/Tests/symm-test-5.stable.out:ws_destroy(ws); pathfinder/tests/Update/Tests/setattr-1.epilogue:ws_destroy(ws); pathfinder/tests/Update/Tests/update.stable.out:ws_destroy(ws); pathfinder/tests/Update/Tests/insert-1.epilogue:ws_destroy(ws); pathfinder/tests/Update/Tests/symm-test-5.epilogue:ws_destroy(ws); pathfinder/tests/Update/Tests/symm-test-3.stable.out:ws_destroy(ws); pathfinder/tests/Update/Tests/symm-test-4.epilogue:ws_destroy(ws); pathfinder/tests/Update/Tests/symm-test-4.stable.out:ws_destroy(ws); pathfinder/tests/Update/Tests/symm-test-3.epilogue:ws_destroy(ws); pathfinder/tests/Update/Tests/symm-test-2.epilogue:ws_destroy(ws); pathfinder/tests/Update/Tests/insert-2.epilogue:ws_destroy(ws); pathfinder/modules/pftijah/pftijah.mx: ws_destroy(ws); pathfinder/compiler/mil/milprint.c: , [m_destroy_ws] = "ws_destroy" pathfinder/compiler/mil/milprint_summer.c: "ws_destroy(ws);\n" ========

Hence, I suppose that either your cvs update / recompile/-install went not completely correct, or you are "by accident" using some other "legacy" code or script when starting the Mserver.

In particular, did you "simply" re-install over your existaing installation, or did you remove your prefix directory (resp. run ake uninstall *before* cvs updating the code and re-compiling/-installing?

I just realise that the former might leave some old modules' .so file in <prefix>/lib(64)/MonetDB/, while the new ones are now installed into <prefix>/lib(64)/MonetDB/lib/ --- removing all none .mil files from <prefix>/lib(64)/MonetDB/ should solve this problem.

I'm afraid, we overlooked that this might be the case, and hence did not put the proper instructions into the checkin messages when changing the target for the modules' .so files from <prefix>/lib(64)/MonetDB/ to <prefix>/lib(64)/MonetDB/lib/. Sorry! Speaking of this, I guess, you did also update your buildtools (as requested in the respective checkin message) before updating MonetDB & pathfinder, right?

Please let us know, whether this helps.

Stefan

On Fri, Oct 13, 2006 at 08:47:54AM +0200, Jan Flokstra wrote:

...

Hi Stefan,

After long idle time I updated and started Pathfinder this morning. I read the discussion about the new ws and I think something is still wrong cleaning up the ws because both in the "MapiClient -lxquery" and with the "xquery()" MIL command I get errors on ws_destroy():

xquery>1+1 <?xml version="1.0" encoding="utf-8"?> <XQueryResult>2</XQueryResult> MAPI = monetdb@localhost:50000 QUERY = 1+1 ERROR = !ERROR: interpret: no matching MIL operator to 'ws_destroy(int)'. !MAYBE YOU MEAN: ! ws_destroy(BAT[void,BAT]) : void xquery>

OR

!WARNING: GDKlockHome: ignoring empty or invalid .gdk_lock. !WARNING: BBPdir: initializing BBP. # Monet Database Server V4.13.1 # Copyright (c) 1993-2006, CWI. All rights reserved. # Compiled for i686-suse-linux/32bit with 32bit OIDs; dynamically linked. # Visit http://monetdb.cwi.nl/ for further information. # Elapsed time = 537us [009us/node] # Shredded 1 XML document (dilbert.xml), total time after commit=0.023s !ERROR: interpret: no matching MIL operator to 'ws_destroy(int)'. !MAYBE YOU MEAN: ! ws_destroy(BAT[void,BAT]) : void

Is this easy to fix??? I can try to fix this or do you want to fix it yourself?

Regards, Jan.