running the regression test001, queries/working_tpch1_compareLogOnly/misc/bug4947.sql always fails with two cpimport crashes. For some reason, it won't fail if run independently.

I've narrowed the cause down to the change from MCOL-498 which allowed for partial filling of files, rather thanfilling with NULLs to the full file size. cpimport, at least for uncompressed tables, uses filesize to determine how much space is left.