On Wed, 31 Jan 2007, Brad Cavanagh wrote:
> Here's one that Tim and I both remember happening in the past, but we can't
> find any record of it, so I'm mailing with hopes that it'll twig some
> memories.
>
> It would seem that FLATCOR doesn't close GLOBAL.sdf. When I run ORAC-DR
> through a batch of files, an NDFTRACE that directly follows a FLATCOR results
> in:
>
> # 63 Err: !! Requested data extends beyond the end of the record; record
> # length is 8
> # 63 Err: ! bytes (possible corrupt HDS container file
> # 63 Err: !
> # /home/bradc/data/oracdr/reduced/uist/20060815/adam_26042/GLOBAL.sdf).
> # 63 Err: ! HDS_OPEN: Error opening an HDS container file.
> # 63 Err: ! SUBPAR: Failed to update GLOBAL file for parameter NDF
> # 63 Err: !! DAT__INCHK: Integrity check
> # 63 Err: Error in obeyw to monolith ndfpack_mon (task=ndftrace): 147358691
>
> If I take the FLATCOR call out, the error message disappears. Does this ring
> any bells? We believe that FLATCOR isn't closing GLOBAL.sdf, but don't want
> to go barking up the wrong tree.
Thanks for the replies, but it's all still really hazy and horribly
annoying to debug. I've tried paring it down to a small group of commands
that can be run in ICL to reproduce the bug, but:
1) the smallest number of commands used by ORAC-DR that still results in
the error in ORAC-DR is 37, and
2) those 37 commands don't reproduce the error in ICL.
Argh.
However, I have determined that it only happens when HDS_64BIT is set to
1, implying an HDS error somewhere.
Brad.
|