Bugs item #1740502, was opened at 2007-06-20 19:59 Message generated for change (Comment added) made by sjoerd You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=482468&aid=1740502&group_id=56967
Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: PF/tests Group: Pathfinder 0.18 >Status: Closed >Resolution: Fixed Priority: 6 Private: No Submitted By: Stefan Manegold (stmane) >Assigned to: Sjoerd Mullender (sjoerd) Summary: PF: live-variable-limit-handling in mil_opt.c buggy Initial Comment: Since Saturday June 16 2007, i.e., after checkins on Friday June 15 2007, pathfinder tests tests/BugTracker/Tests/wrong_result_duplication.SF-1549292.xq and tests/XQuery/Tests/transient_upward_steps.xq fail with ERROR = !ERROR: no matching MIL operator to 'leftfetchjoin(BAT[oid,void], BAT[oid,oid])'. and ERROR = !ERROR: no matching MIL operator to 'insert(BAT[oid,str], void, str)'. respectively; cf. http://monetdb.cwi.nl/testing/projects/monetdb/Stable/pathfinder/.mTests103/GNU.64.64.d-Fedora6/tests_BugTracker/wrong_result_duplication.SF-1549292.err.00.html and http://monetdb.cwi.nl/testing/projects/monetdb/Stable/pathfinder/.mTests103/GNU.64.64.d-Fedora6/tests_XQuery/transient_upward_steps.err.00.html ---------------------------------------------------------------------- >Comment By: Sjoerd Mullender (sjoerd) Date: 2007-07-20 15:11 Message: Logged In: YES user_id=43607 Originator: NO Fixed in CVS. There was indeed a problem when there were too many variables. No new test needed. ---------------------------------------------------------------------- Comment By: Stefan Manegold (stmane) Date: 2007-07-18 21:19 Message: Logged In: YES user_id=572415 Originator: YES The problem seems to be triggered by this checkin: ======== $ cvs log -N -r1.23.2.3 compiler/include/mil_opt.h RCS file: /cvsroot/monetdb/pathfinder/compiler/include/mil_opt.h,v Working file: compiler/include/mil_opt.h head: 1.25 branch: locks: strict access list: keyword substitution: kv total revisions: 42; selected revisions: 1 description: ---------------------------- revision 1.23.2.3 date: 2007/06/15 12:45:48; author: boncz; state: Exp; lines: +3 -3 - now that I think of it, reduce by a further factor 16 (256KB) ============================================================================= $ cvs diff -r1.23.2.{2,3} compiler/include/mil_opt.h Index: compiler/include/mil_opt.h =================================================================== RCS file: /cvsroot/monetdb/pathfinder/compiler/include/mil_opt.h,v retrieving revision 1.23.2.2 retrieving revision 1.23.2.3 diff -u -r1.23.2.2 -r1.23.2.3 --- compiler/include/mil_opt.h 15 Jun 2007 12:35:24 -0000 1.23.2.2 +++ compiler/include/mil_opt.h 15 Jun 2007 12:45:48 -0000 1.23.2.3 @@ -24,7 +24,7 @@ * 2000-2005 University of Konstanz and (C) 2005-2007 Technische * Universitaet Muenchen, respectively. All Rights Reserved. * - * $Id: mil_opt.h,v 1.23.2.2 2007/06/15 12:35:24 boncz Exp $ + * $Id: mil_opt.h,v 1.23.2.3 2007/06/15 12:45:48 boncz Exp $ */ #include <stdio.h> @@ -45,9 +45,9 @@ */ #define OPT_STMTS 32767 /* <65535 if first-use further out than this amount of stamements, dead codes survises anyhow */ -#define OPT_VARS 1024 /* don't try dead code elimintation above this amount of live variables */ +#define OPT_VARS 256 /* don't try dead code elimintation above this amount of live variables */ #define OPT_REFS 63 /* keep track of usage dependencies; may omit some, which results in surviving dead code */ -#define OPT_CONDS 1024 /* maximum if-nesting */ +#define OPT_CONDS 256 /* maximum if-nesting */ #define OPT_SEC_PROLOGUE 0 #define OPT_SEC_QUERY 1 ======== re-increasing OPT_VARS to (e.g.) 512 make the tests work correctly, again; hence, I guess there is a bug in the live-variable-limit-handling in the MIL code optimizer. Unfortunately I have not (yet?) managed to locate the actual problem... ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=482468&aid=1740502&group_id=56967 ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ Monetdb-bugs mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/monetdb-bugs
