Bugs item #1637825, was opened at 2007-01-17 16:53 Message generated for change (Comment added) made by tsheyar You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=482468&aid=1637825&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 CVS Head >Status: Closed >Resolution: Fixed Priority: 8 Private: No Submitted By: Stefan Manegold (stmane) Assigned to: Jan Rittinger (tsheyar) Summary: PF: XRPC tests fails Initial Comment: While working fine until Tuesday Jan 9 2007, all XRPC tests fail since (at least) Friday Jan 12 2007 (there were no pathfinder testing results in between these dates), i.e., after checkins on Tuesday Jan 9 2007 through Thursday Jan 11 2007 (which include propagations of changes from the XQuery_0-14 branch that were originally applied there between Friday Jan 05 2007 and Thursday Jan 11 2007). fatal error: The Pathfinder compiler experienced an internal problem in the MIL code generation. You may want to report this problem to the Pathfinder development team ([EMAIL PROTECTED]). When reporting problems, please attach your XQuery input, as well as the following information: could not lookup variable '#pf:v_0219' in join recognition. We apologize for the inconvenience... See (e.g.): http://monetdb.cwi.nl/testing/projects/monetdb/Current/pathfinder/.mTests103/GNU.64.64.d-Fedora4/tests_XRpc/add_dbl.err.00.html http://monetdb.cwi.nl/testing/projects/monetdb/Current/pathfinder/.mTests103/GNU.64.64.d-Fedora4/benchmarks_XMark_XRpc/q01.err.00.html ---------------------------------------------------------------------- >Comment By: Jan Rittinger (tsheyar) Date: 2007-01-19 10:41 Message: Logged In: YES user_id=993208 Originator: NO The bug is now fixed in CVS. Thanks Jennie for hunting this one down. The bug was hidden in coreopt.brg. It was however not related to my latest changes but only triggered by this rewrite. The variable expansion that removed the binding did not fill in the replacement because the XRPC operator rule did use the wrong token (var instead of Atom). ---------------------------------------------------------------------- Comment By: Ying Zhang (yingying) Date: 2007-01-19 00:12 Message: Logged In: YES user_id=341633 Originator: NO This problem is caused by Jan R's checkin on Jan. 09: ==================== From: Jan Rittinger <[EMAIL PROTECTED]> Date: Tue, 09 Jan 2007 14:40:58 +0000 Update of /cvsroot/monetdb/pathfinder/compiler/core In directory sc8-pr-cvs7.sourceforge.net:/tmp/cvs-serv4167/core Modified Files: coreopt.brg fs.brg Log Message: ... -- Introduced new core optimizations: * casts are removed whose input type already matches the cast type ... @@ -1067,6 +1085,16 @@ } break; + /* CoreExpr: cast (SequenceType, CoreExpr) */ + case 72: + if (PFty_subtype (TY(L(p)), TY(R(p))) && + PFty_subtype (TY(R(p)), TY(L(p)))) { + *p = *R(p); + rewritten = true; + break; + } + break; + /* Atom: var */ case 73: { ==================== I used the following query and module: ==================== $ cat add_dbl.xq import module namespace xrpc = "xrpc-fun" at "/ufs/zhang/www/xrpc/xrpcfun.xq"; for $i in (40.5, 50.5) return execute at {"localhost"} {xrpc:add($i, (10.5, 30.5))} $ cat /ufs/zhang/www/xrpc/xrpcfun.xq module namespace xrpcfun = "xrpc-fun"; declare function xrpcfun:add($v1 as xs:decimal, $v2 as xs:decimal*) as xs:decimal* { let $res := $v1 for $i in $v2 return $res + $i }; ==================== The pathfinder compiler gives the following error: $ pf add_dbl.xq fatal error: The Pathfinder compiler experienced.... could not lookup variable '#pf:v_0041' in join recognition. ... # halted in ../../../compiler/mil/milprint_summer.c (mps_error), line 158 The variable v_0041 should contain the URI string of the destination of an XRPC call, but its value is removed during core optimization. Before the call to reduce() in PFcoreopt() (coreopt.brg line 1190), a fraction of the XQuery Core tree looks like this: ... flwr (nil (), flwr (let (letbind (var (#pf:v_0040), apply (fn:data, arg (lit_str (localhost), nil ()))), let (letbind (var (#pf:v_0041), cast (seqtype (string), var (#pf:v_0040))), nil ())), xrpc (var (#pf:v_0041), ... After the call, the XQuery Core tree look like this: ... flwr (let (letbind (var (#pf:v_0022), seq (lit_dec (40.5), lit_dec (50.5))), for (forbind (forvars (var (i), nil ()), var (#pf:v_0022)), nil ())), xrpc (var (#pf:v_0041), flwr (let (letbind (var (#pf:v_0032), seq (lit_dec (10.5), lit_dec (30.5))), nil ()), apply (xrpcfun:add, arg (var (i), arg (var (#pf:v_0032), nil ()))))))) ... Obviously, the value of v_0041 was removed during the optimization, by the new "case 72". The simplest solution is to remove "case 72", and all XRPC tests work again, but it might not be Jan's intension... Jan, I reassign this bug to you. Would you please have a look at this to see if the implementation of "case 72" can be improved? Otherwise, would you mind to (temporarily) undo this change? Thank you very much! Jennie ---------------------------------------------------------------------- Comment By: Stefan Manegold (stmane) Date: 2007-01-17 18:20 Message: Logged In: YES user_id=572415 Originator: YES Jennie, maybe you could have a look at this. Feel free to assign it to someone else, in case you think it's not your bug. Thanks! ---------------------------------------------------------------------- Comment By: Stefan Manegold (stmane) Date: 2007-01-17 17:51 Message: Logged In: YES user_id=572415 Originator: YES Since these tests do work in the XQuery_0-14 release branch, i.e., they have (recently) been broken only in the development trunk, I consider this bug report high priority. ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=482468&aid=1637825&group_id=56967 ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys - and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ Monetdb-bugs mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/monetdb-bugs
