[daip] Re: AIPS bug ($QOOP/OUVSRT)

[Eric Greisen]

Bill Cotton writes:
 > Eric,
 >    Your OUVSRT blows up when passed compressed data.  In principle, it
 > could work OK without decompressing; however, SRTIN gets an incorrect
 > value of LREC from common (appropriate for uncompressed data) and the
 > lower level sorting routines choke trying to interprete compressed
 > data as random parameters. 
 > 

I think that I have found a variety of factors entering here:

1. OUVCLN  decides about the compression state of its output based
   solely on the keyword in the input DOUVCOMP.  If that is false or
   missing, the output is uncompressed.

2. In QUVUTIL subroutine OUVSRT gets the input header, then closes the
   input and closes the output.  It was using CATBLK and the close of
   the output replaced CATBLK with the output header.  This CATBLK was
   then passed to UVSORT which calls UVPGET pretty directly and had
   the correct LREC for the output file.  If that is wrong for the
   input file, it is because of error 1 above (or some other failure
   to match the 2 compression states).  I am adding a defense of the
   input header but also a test that the input and output compression
   states are the same in OUVSRT.

Neither of these is the source of your problems with CC filtering -
the messages there suggest that OUVSRT was never called and that the
error arose in the routine that tries to decide wheter to call it.

I guess that we could change #1 above also.  But I bet that breaks a
lot of stuff, so I prefer not to do so.

I will putbck my revised QUVUTIL (which has your changes in it) and
QCLEAN after I fugure out how to get an FQ table copied too.  UVDTCO
copies AN happily but does not do FQ.

Eric