Tim, Six & a half thoughts:
1. You say 1:1 correspondence for ID:PHYS.PALLET.ID, meaning no multivalues, right? I ask about multivalues because if there are repeated identical mv entries in a record, SELECTINDEX performs differently in PI vs PICK flavours. See $OPTIONS PIOPEN.SELIDX. 2. And no null PHYS.PALLET.IDs, right? 3. It is possible to corrupt an index if the primary data file were copied using an OS-level utility so that the header in the copy still contains pointers to the index. for example you CTRL+[mouse right click] and drag a copy to your sandbox to test with. To search for all indexes, in each VOC do: SELECTF SETINDEX INFORM 4. Corruption can occur if there is a VOC pointer to the path of the type-25 index file itself. (i.e.,, VOC <2> will look something like ...\I_[datafilename]\INDEX.004). That allows you to edit or otherwise manipulate the index valued directly. For that matter basic's OPENPATH could do the same w/o a VOC reference. 4a. You might want to ~temporarily~ create such a VOC item so you can examine the contents of the index just for troubleshooting. 5. If you rebuild the index does it work right at first, then somehow get corrupted? Or is it always wrong? 6. TCL SELECT works correctly? only Basic's SELECTINDEX fails? How about BSCAN? Beyond that, maybe you could post an example of where it fails? (w/o =0D=OA's?) cds ------- u2-users mailing list u2-users@listserver.u2ug.org To unsubscribe please visit http://listserver.u2ug.org/
