[daip] problems with flag table

[Eric Greisen]

J.H. van Gorkom writes:
 > FLGIT OUTFGVER=   3  / output flag table version
 > FLGIT NFGREC=   1692060  / flag table records written
 > FLGIT / Flagged      5931 channels for input flux > clip level 5.000E+01
 > FLGIT / Flagged         0 channels for VPOL flux > clip level  0.000E+00
 > FLGIT / Flagged         0 channels for residual fit flux >     1.500E+00
 > FLGIT / Flagged   3631554 channels for residual signal flux >  2.000E+00
 > FLGIT / Flagged      1407 channels for residual flux >  6.00*RMS(f)
 > FLGIT / Flagged     25777 channels for residual Re/Im >  5.00*RMS(R/I)
 > FLGIT / Flagged         0 channels previously or in calibration
 > FLGIT / 0.0137 of the channels are now flagged
 > FLGIT / Full spectra flagged:       0
 > 
 > Note that only 1% of the channels is flagged, despite the HUGE number of 
 > flags.

     Note that the flagging was able to take account of consecutive
channels etc and so ONLY made 1.7 million while flagging actually 3.7
million.

 > We then tried UVCOP with flagver 3 and uvcopprm -1, 0
 > I understand that it should drop flagged visibilities in that
mode.

    only if all channels and ifs and stokes for a particular time are
deleted

 > 
 > HERCUL> UVCOP1: Task UVCOP  (release of 31DEC07) begins
 > HERCUL> UVCOP1: Create A963        .UVCOP .   1 (UV)  on disk 10  cno   39
 > HERCUL> UVCOP1: Copied from the beginning
 > HERCUL> UVCOP1:         to  the end
 > HERCUL> UVCOP1: Copied    65429. vis records
 > HERCUL> UVCOP1: Dropped       0. flagged vis records
 > HERCUL> UVCOP1: Updating tables for IF/FREQID/channel selection
 > HERCUL> UVCOP1: Copied AN file from vol/cno/vers 10   19   1 to 10   39   1
 > HERCUL> UVCOP1: Copied CL file from vol/cno/vers 10   19   1 to 10   39   1
 > HERCUL> UVCOP1: Copied CL file from vol/cno/vers 10   19   2 to 10   39   2
 > HERCUL> UVCOP1: Copied SU file from vol/cno/vers 10   19   1 to 10   39   1
 > HERCUL> UVCOP1: Appears to have ended successfully
 > HERCUL> UVCOP1: hercules02   31DEC07 TST: Cpu=      20.6  Real=  160

        Did you delete flag tables 1 and 2?  If not, there should be a
copied message here for UVCOP to have copied them - it omits messages
for mising tables and those that it applied.
 > 
 > It did not drop the flagged data as a UVPLT shows me.
 > 
 > CLIP produces a similar flag table and has the same problem.
 > 

I have tried making an enormous number of flags also and then running
UVCOP with both a GNU-compiled version and the Intel-compiled version
we ship in the binary ditribution and there was no problem.  Do you
compile AIPS locally or use the binary distribution?  If the former,
what version of the GNU compiler did you use (or some other compiler)?

Eric Greisen