On Tuesday 29 May 2007 15:38, Stefan Manegold wrote:

On Tue, May 29, 2007 at 12:20:21PM +0000, Jan Flokstra wrote:

...

Update of /cvsroot/monetdb/pathfinder/modules/pftijah In directory sc8-pr-cvs16.sourceforge.net:/tmp/cvs-serv10231

Modified Files: pftijah.mx Log Message: - try to repair pftijah "testcoll2" bug on 64bit archs

Index: pftijah.mx =================================================================== RCS file: /cvsroot/monetdb/pathfinder/modules/pftijah/pftijah.mx,v retrieving revision 1.123 retrieving revision 1.124 diff -u -d -r1.123 -r1.124 --- pftijah.mx 25 May 2007 12:34:24 -0000 1.123 +++ pftijah.mx 29 May 2007 12:20:18 -0000 1.124 @@ -3267,15 +3267,16 @@ return GDK_FAIL; }

- doc_start = *(oid*)BUNtail(doc_firstpre,r); - oid tj_nextIndex = tj_docIndex + 1; + // 64bit ERROR?? doc_start = *(oid*)BUNtail(doc_firstpre,r);

^^^^^^^^^^^^^ No, a bug in the code: BAT doc_firstpre is [oid,int]., i.e., its tail is int, not oid, and int != oid; ========

We were aware of this tail type and that's why we created the cast. Henning was a bit hesitant to change the bat type to [oid,oid] because he expected problems with '<' and '>' so we decided to create the extra cast. The value can never be negative so we never checked for this. We could make it an assert(); Furthermore the text //64bit ERROR was to remind us of our own error on 64bit oid arch's and not an error of monetdb. JanF.

pathfinder/modules/pftijah/pftijah.mx-73-.COMMAND pf2tijah_node( pathfinder/modules/pftijah/pftijah.mx-74- BAT[oid,str] doc_name, pathfinder/modules/pftijah/pftijah.mx:75: BAT[oid,int] doc_firstpre, ^^^^^^^^^^^^^^^^^^^^^^^^^ pathfinder/modules/pftijah/pftijah.mx-76- BAT[oid,oid] pfpre, pathfinder/modules/pftijah/pftijah.mx-77- BAT[oid,oid] item, pathfinder/modules/pftijah/pftijah.mx-78- BAT[oid,int] kind, pathfinder/modules/pftijah/pftijah.mx-79- BAT[oid,str] doc_loaded) pathfinder/modules/pftijah/pftijah.mx-80- : BAT[void,oid] = CMDpf2tijah_node; pathfinder/modules/pftijah/pftijah.mx-81- "Translate Pathfinder node sequence to tijah node sequence" ======== hence, to access the tail in C, you (obviously!) need to use int doc_start = *(int*) BUNtail(doc_firstpre,r); ^^^ ^^^ if (for what ever reason) doc_start needs to be of type oid instead of int, you need to use oid doc_start = (oid) *(int*) BUNtail(doc_firstpre,r); ^^^ ^^^ ^^^ (as you do below); in fact, you should first check, whether the respective tail value of doc_firstpre is not negative, before you caat it to oid!

...

+ doc_start = (oid)*(int*)BUNtail(doc_firstpre,r); + oid tj_nextIndex = tj_docIndex + (oid)1; if ( BATcount(doc_firstpre) > tj_nextIndex ) { r = BUNfnd(doc_firstpre,&tj_nextIndex); if ( !r ) { stream_printf(GDKout,"Cannot do range for tijah-firstpre @ %d.\n",tj_docIndex); return GDK_FAIL; } - doc_end = *(oid*)BUNtail(doc_firstpre,r) - 1; + doc_end = *(oid*)BUNtail(doc_firstpre,r) - (oid)1;

NOPE! either int doc_end = *(int*) BUNtail(doc_firstpre,r) - 1; or oid doc_end = (oid) (*(int*) BUNtail(doc_firstpre,r) - 1); plus a check, whether *(int*) BUNtail(doc_firstpre,r) is > 0 !

Stefan

...

} else { doc_end = oid_nil; }

------------------------------------------------------------------------- This SF.net email is sponsored by DB2 Express Download DB2 Express C - the FREE version of DB2 express and take control of your XML. No limits. Just data. Click to get it now. http://sourceforge.net/powerbar/db2/ _______________________________________________ Monetdb-pf-checkins mailing list Monetdb-pf-checkins@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-pf-checkins