[daip] VBGLU in 2007 AIPS
Eric Greisen
egreisen at nrao.edu
Thu Oct 29 12:44:37 EDT 2009
Geoffrey Bower wrote:
> Craig, Eric, et al,
>
> Yuri Kovalev just told me that VBGLU in 31 DEC 2007 AIPS has a bug that
> affects 512 Mb/s data correlated in two passes. Is this true? I have
> never heard anything of this kind. This obviously would have
> significant impact on our RIPL processing done with 2007 AIPS.
>
> Cheers,
>
> Geoff
12777. December 26, 2008 VBGLU Eric
The task had 3 main problems:
1. When disk I/O had to be restarted because the required UV
record preceded the ones in RAM, the task checked a few of the
upcoming records to find the minimum needed input record
number. Thus the first record number in the new buffer could
be one that preceded the currently required one.
Unfortunately, the pointer was set to the first in the new
buffer always. This led to wrong baseline data being included
in the output record. Raised the search region and made sure
it did not search over empty records and made the pointer aim
at the desired record.
2. On AT tables, the task asks for a SOURCE column that is not
present in the table and so quit on AT tables. Added a
spexcial error test for this, allowing it to procede normally
since the copy can handle a missing selection column.
3. The handling of character-data columns was incorrect. So
far as I can tell, the error only damaged CQ tables.
Corrected all the length handling for character-valued columns
of 1 or more dimensions (includes AT, CT, and SU tables).
Moved to 31DEC08 this date.
This was discovered nearly at the end of the 08 version. It is point 1
above that could really mess up things.
There were fixes in 07 too but if your version was complete to the end
of that release then they are not relevant.
Note that the aips documentation web pages have a tool to extract info
from CHANGE.DOC for you by task name.
ERic Greisen
More information about the Daip
mailing list